Skip to main content

Modelling properties, interactions, toxicity and environmental behaviour of engineered nanoparticles

Final Report Summary - NANOPUZZLES (Modelling properties, interactions, toxicity and environmental behaviour of engineered nanoparticles)

Executive Summary:
Some types of engineered nanoparticles (NPs) can be toxic for living organisms and exhibit negative impact on the environment. Thus, the design of new nanomaterials must be supported by a rigorous risk analysis. The main objective of the NanoPUZZLES project is to create new computational methods for comprehensive modelling of the relationships between the structure, properties, molecular interactions and toxicity of engineered nanoparticles. The following scientific project achievements can be highlighted:
• Standardised data collection templates, based upon ISA-TAB-Nano, for capturing important toxicity and physicochemical/structural measurements, as well as key experimental details, were developed.
• Novel approaches for scoring the quality of nanomaterial physicochemical/structural and toxicity data were established.
• A spreadsheet of 336 annotated primary literature references, reporting nanomaterial data for nanosafety relevant endpoints (principally the prioritised endpoints of cytotoxicity, genotoxicity and aquatic toxicity, as well as cellular uptake data), was completed and made publicly available via the Zenodo (http://dx.doi.org/10.5281/zenodo.35419) and FigShare (https://figshare.com/search? q=NanoPUZZLES+project&quick=1) online, searchable resources.
• New algorithms for calculating the descriptors i.e. parameters that characterize the unusual structure of NPs and electronic states resulting from quantum effects of the nano-size of investigated materials have been developed.

Within NanoPUZZLES project five types of descriptors potentially useful in developing new predictive models for the activities and properties of the nanoparticles such as:
(1) topological descriptors,
(2) descriptors derived from quantum-mechanical calculations,
(3) descriptors derived from computational processing of microscopic (SEM/TEM/AFM) images,
(4) descriptors based on “anisotropy dimensions” and
(5) descriptors based on graph or fractal theory were evaluated. In addition we have introduced three novel kinds of descriptors (not previously anticipated in the DoW) i.e.: (1) «Liquid Drop» Model (LDM) descriptors, (2) zero-dimensional (0D) - Constitutional descriptors and (3) Simplex Representation of Molecular Structure (SiRMS) descriptors.
• A computational protocol (workflow) for the calculation of the properties of large interacting systems, involving nanoparticles (NPs) and biological molecules (BMs) was developed using atomistic and electronic models. Specific rules (workflows) have been proposed for the computation of the interaction properties of small, medium and large size systems. In small interacting systems the Energy Decomposition Analysis (EDA), developed by Su and Li in connection with DFT (e.g. B3LYP/3-21G*) was recommended, whereas molecular dynamics (MD) gives credible results for the binding free energy (MM#PBSA) of large interacting systems.
• The potential benefits of using chemoinformatics approaches such as Nano-QSAR and Nano-QSPR modelling to obtain predictive knowledge for organic and inorganic nanoparticles that affect human cells and environment, and utilize this knowledge to improve the experimental design of NPs and enable their prioritization for in vivo testing have been demonstrated. Based on experimental data from WP1 and novel nanodescriptors calculated within WP2 we have demonstrated that by applying Nano-QSAR and Nano-QSPR models the biological activity (e.g. cytotoxicity, mutagenicity, genotoxicity, cellular uptake, membrane damage) and phys/chem properties (e.g. thermal conductivity, viscosity, dispersibility in various solvents) of various nanoparticles (i.e. metal oxides nanoparticles, fullerenes, multi-walled carbon-nanotubes, nanofluids, graphene as well as superparamagnetic iron oxide nanoparticles (SPIONs) decorated with organic molecules) could be estimated with a similar level of accuracy as provided by experiments.
• Novel and effective algorithm for filling data gaps in quantitative manner (read-across approach) was introduced. In addition conceptual framework for further grouping of NPs was established.
• The usefulness of methods of causal discovery to elucidate the underlying structure of the nanotoxicity data and retrieve additional, more robust interpretation for the developed SAR models was demonstrated. We have presented that the causal structures can efficiently be used in Nano-SAR modelling as additional criteria for quality evaluation.
Project Context and Objectives:
Nanotechnology is a rapidly expanding area of research with huge potential in many sectors. Different types of nanoparticles (NPs) find a vast range of medicinal applications. However, some types of engineered nanoparticles can be toxic for living organisms and exhibit negative impact on the environment. Thus, the design of new nanoparticles must be accompanied by a rigorous risk analysis. Following the European recommendations and bearing the ethical aspects of such research in mind, the risk assessment procedures should be performed with possible reduction of animal testing. One of the most promising alternatives is the application of computational techniques, which not only allows curtailing animal use, but also enables a significant reduction of in cost of the required risk assessment.

Consequently, the research conducted within the “Modelling properties, interactions, toxicity and environmental behaviour of engineered nanoparticles” project (acronym: NanoPUZZLES) is aimed at developing, within three years, a package of computational algorithms for the comprehensive modelling of the relationships between the structure, properties, molecular interactions and toxicity of selected classes of engineered nanoparticles (NPs).
The package
(i) will serve as a proof-of-concept that the risk related to NPs can be comprehensively assessed with use of computational techniques and
(ii) will define a basis for development of further modelling techniques for a large variety of nanoparticles.

Computational algorithms and methodological approaches will be developed within four work packages related to the following thematic areas:
• Quality assessment of physicochemical and toxicological data available for nanomaterials and data exploration (WP1: NanoDATA),
• Development of novel descriptors for nanoparticles’ structure (WP2: NanoDESC),
• Simulating interactions of nanoparticles with biological systems (WP3: NanoINTER),
• Quantitative and qualitative structure-activity relationship modelling, grouping and read across (WP4: NanoQSAR).

The NanoPUZZLES project objectives, as defined in the Description of Work are listed below along with the deliverables (as defined in Annex I to the Grant Agreement):

WP1: NanoDATA
The main objective of the first thematic area is to develop a framework for classifying engineered nanoparticles based on the existing data utilising pattern recognition methods. This includes creating knowledge of engineered nanoparticles (NPs) by:
• Defining and compiling high quality chemical structure for the NPs
• Associating, where possible, curated and quality controlled toxicity data to structures
• Development of a searchable inventory of knowledge within a publicly available database
• Forming groups of engineered nanoparticles based on the structure and properties data to allow for their categorisation.

WP2: NanoDESC
The main objective of the second thematic area is to develop a framework to optimally characterise the structure of engineered NP using appropriate descriptors and categorising them according to structural similarities.
This includes:
• Evaluation of the existing systems currently available for structural characterization of NPs.
• Development of simplified molecular models sufficient to characterize the whole structure
• Development of descriptors for the nanostructure (“nano-descriptors”) of four types:
1) topological descriptors, which are calculated with molecular graphs, SMILES, InChI, SMART notations, and descriptors based on the technological and physicochemical parameters,
2) descriptors derived from quantum-mechanical calculations,
3) descriptors derived from computational processing of microscopic (SEM/TEM/AFM) images of the particles,
4) descriptors based on the anisotropy dimensions (as proposed by Glotzer and Solomon).
• Development of databases of the physicochemical and biochemical properties of the nanomaterials which will be made available via the internet
• Development of freeware for calculating nanodescriptors.

WP3: NanoINTER
The objective of the third thematic area is to develop a computational protocol to satisfactorily predict and explain interactions between engineered NPs and biological systems as well as small molecules.
This includes:
• Development of a protocol, which will provide the guidelines for developing or implementing a model for the study of large interacting systems.
• Development of a hierarchy of computational models for the study of interacting systems involving NPs and biological molecules of varying size.
• Development of techniques for the study of the environment (e.g. solvent) on the interacting system.
• Study of the effect of the computational model (e.g. level of quantum-mechanical theory) on the results.
• Implementation of techniques for the resolution of the interaction energy into various contributions (e.g. those due to electrostatic forces, dispersion etc).
• Design/recognition of functional groups, which seriously reduce the genotoxicity and increase the solubility of the considered NPs.
• Study of factors affecting the interaction of the selected systems, nanoparticles (NP) /{biological molecule}.

There are several important factors related with the NP. Among those we note:
(i) the chemical composition of the NP (e.g. fullerene, CNT, etc.);
(ii) the size and shapeof the NP;
(iii) the particle aggregation;
(iv) the surface charge of the NPs, which is known to affect their cellular uptake;
(v) contamination. NPs (e.g. CNTs) may involve one or more toxic metals (e.g. Fe, Co, Ni) which may be considered as contaminants;
(vi) functionalization. It is understood that functionalization may affect the toxicity of the NP as well as its solubility. We shall look for functional groups, which seriously reduce the genotoxicity and increase the solubility of the considered NPs. Thus we propose to consider how the above factors affect the interaction of the selected systems: NP/{biological molecule}. Moreover, engineered nanoparticles exposed to environment participate in reactions of other environmental pollutants (oxidation reactions etc.) and can change as reaction rates of degradation (oxidation) processes of those pollutants, as well as to change reaction pathways and produce new metabolites.

WP4: NanoQSAR
The objective of this thematic area is to develop scientifically justified and technically viable methods to quantitatively model the relationships between chemical structure and toxicological targets as well as to extend the understanding of toxicity and behaviour of emerging nanoparticles by establishing relations between experimental (based on available, validated data) and computational properties.
This includes:
• Investigating the impact of size on the physico/chemical properties of NPs at the appropriate level of the quantum-mechanical theory.
• Developing NanoQSAR models of toxicity and environmentally relevant physico/chemical endpoints, based on reliable experimental data and appropriate nano-descriptors.
• Comparing the efficiency of CoMFA/CoMSIA and Hansch Analysis modelling schemes in NanoQSAR.
• Investigating the minimum requirements sufficient for successful validation of NanoQSAR models (minimal number of data, evaluation of the applicability domain etc.) in the light of the OECD Principles for the Validation of (Q)SARs.
• Development of procedures for validating QSPR/QSAR models using probabilistic principles: balance of correlations, balance of correlations with ideal slopes, and filtration of the rare attributes, which can lead to overtraining.
• Estimating the environmental behaviour of NPs based on the physico/chemical data predicted with NanoQSAR.
• Evaluation and publication of the NanoQSAR models and the results in scientific journals and with use of QSAR reporting formats (QMRFs) and QSAR prediction reporting formats (QPRFs).
• Update of the database (including the predicted results).
• Development of the conceptual framework for further grouping NPs based on chemical structure, physicochemical properties, interactions and toxicity.

This part of the NanoPUZZLES brings together all findings and summarizes the results of the project. High quality experimental physicochemical and toxicological data (from Work Package 1: NanoDATA) and novel descriptors of nanostructure (developed in Work Package 2: NanoDESC) will be utilized to develop mathematical models describing relationships between the structure and properties/activity. Information on the significance of structural factors responsible for the observed activity will be delivered by Work Package 3: NanoINTER. The information about the character of interaction mechanisms will be important for an appropriate selection of nanodescriptors representing structural features of the studied nanoparticles.

Application of the methods developed within the four thematic areas will allow for predicting toxicity and the behaviour of novel nanoparticles from their structure and/or physicochemical properties without the necessity of performing extensive empirical testing (reduction of costs and need for animal testing). Moreover, it will result in a framework being established to categorise nanoparticles according to the potential for exposure, as well as physicochemical, structural and toxicological properties (based on available empirical data and computationally predicted results). This, in the longer perspective, should lead to designing and engineering nanomaterials that are of low risk for human and the environment.

In addition to the four research WPs (WP1-WP4), two additional work packages were planned:
• WP5: Dissemination,
• WP6: Management

WP5: Dissemination
The objective of this work package is to disseminate the project results to relevant stakeholders.

WP6: Management
The objective is to ensure successful implementation of the project plan on time and at cost.
Project Results:
The main objective of the NanoPUZZLES project was to create new computational methods for comprehensive modelling of the relationships between the structure, properties, molecular interactions and toxicity of engineered nanoparticles. Direct collaboration between NanoPUZZLES project partners and their contribution to the different tasks, as well as research cooperation and integration activities among other projects within the NanoSafety Cluster resulted in the following scientific project achievements:

WP1: NanoDATA:

Task 1.1: Data Collection

Initial literature derived datasets were obtained via (a) interactions with the NanoBRIDGES project and (b) additional work within NanoPUZZLES. These initial datasets were organised within non-standardised Excel workbooks and few experimental details were extracted. Following discussions with other NanoSafety Modelling Cluster projects (i.e. MembraneNanoPart, ModEnpTox, MODERN, PreNanoTox, NanoPUZZLES, ModNanoTox and NanoTransKinetics) in June 2013, it was decided to prepare the final NanoPUZZLES datasets according to the ISA-TAB-Nano specification, which was proposed as a global data exchange standard by its developers. The use of a standard data model was recognised as an important enabling step for achieving the objective of submitting the data to a “searchable inventory of knowledge within a publicly available database”. Furthermore, ISA-TAB-Nano supports links to terms from ontologies and the use of standardised terms was recognised as important in facilitating integrated analysis of the final NanoPUZZLES datasets as well as interoperability with datasets/databases from other projects.

A thorough, critical assessment of ISA-TAB-Nano was carried out to make sure that the specification was used appropriately within NanoPUZZLES and that any problems with the existing version of the specification were resolved. This assessment involved providing feedback to the developers of ISA-TAB-Nano in the US Nanotechnology Working Group as well as discussions with the MODERN and ModEnpTox EU projects. In addition, case studies were carried out in order to evaluate the suitability of ISA-TAB-Nano, as it existed at the time for organising nanomaterial data and to determine how to best use it within NanoPUZZLES. These case studies entailed thorough collection of data, including details regarding experimental conditions and techniques, from primary literature articles.

It was determined that various challenges were associated with the use of the evaluated version of ISA-TAB-Nano, including ambiguity perceived in the existing documentation and difficulties associated with creating appropriately named columns: creating column names using standardised terms from ontologies at the point of data collection was deemed inefficient and the potential need for new ontology terms to be defined was identified. Proposals were developed for resolving these challenges, such as the creation of specific templates based upon the ISA-TAB-Nano specification with pre-defined columns to be used in ongoing data collection efforts. Additional challenges facing ongoing data collection efforts included the need to define appropriate minimum information standards, as well as potential copyright restrictions, which could affect, for example, the inclusion of transmission electron microscopy (TEM) images in the publicly released data collection. These challenges and possible solutions were discussed amongst NanoPUZZLES project partners as well as with both the NanoSafety Modelling Cluster and the NanoSafety Cluster Databases Working Group.

A detailed description of the initial data collection activities referred to above was presented in deliverable D1.1. Subsequent data collection activities, within the first reporting period, are summarised below.

Following discussions between all NanoPUZZLES project partners in December 2013, it was decided to prioritise the collation of genotoxicity and cytotoxicity data. A list of articles with nanomaterial genotoxicity and/or cytotoxicity data was identified via evaluating over 250 articles reported in the initial datasets or retrieved via additional literature searches. In total, 165 primary references were identified with relevant data for one or more nanomaterials. The evaluation of these references included an initial assessment of their usefulness for building nano-QSAR models: the availability of genotoxicity and/or cytotoxicity data, the number of nanomaterials studied and the availability of electron microscopy images (required by WP2 and WP4) was summarised for each reference. The summaries of the data contained within all these articles serve as a basis for prioritising articles from which data are being extracted into ISA-TAB-Nano files.

In order to avoid duplication of effort, articles which were identified by the Harmonisation of Modelling Projects initiative were excluded from this list and a Google spreadsheet was distributed to all participating projects indicating which references were being used to derive the final datasets/databases being worked on within each project. Please note that this initiative involves the following projects: MembraneNanoPart, MODERN, ModEnpTox, PreNanoTox, NanoPUZZLES.
Since the ISA-TAB-Nano specification, as proposed by its developers, does not specify exactly which nanomaterial characteristics, experimental details and measurements should be recorded, data collection templates, based upon ISA-TAB-Nano, were designed within NanoPUZZLES to capture important toxicity and physiochemical/structural measurements as well as key experimental details. The experimental details/measurements, which these templates were designed to capture were identified based on consulting the literature, the MODENA Cost Action, the PreNanoTox database and active participation in discussions within the NanoSafety Cluster Databases Working Group. These templates were developed by extending the generic templates available from the US Nanotechnology Working Group. A more detailed description of the NanoPUZZLES data collection templates was provided in deliverables D1.2 and D1.3.

These data collection templates have been used to record data extracted from prioritised journal articles and scientific reports identified via the literature analysis described above. Extraction of data from additional prioritised articles will continue in the next reporting period.

In parallel to populating the ISA-TAB-Nano based data collection templates, additional data were obtained via interactions between NanoPUZZLES WP1 and external parties as well as further literature research. An export of data from the Nanomaterial-Biological Interactions (NBI) Knowledgebase (http://nbi.oregonstate.edu/) was provided upon request by Dr Stacey Harper and Bryan Harper. Additional data were extracted from the literature as part of the Harmonisation of Modelling Projects initiative and a session was arranged to provide feedback to the developers of the PreNanoTox database into which data were entered as part of this initiative. Recent discussions within NanoPUZZLES led to aquatic toxicity, an important ecotoxicological endpoint, being added to the set of prioritised endpoints and fullerene genotoxicity data was required by WP3. Hence, new literature searches identified additional articles with fullerene genotoxicity data or Daphnia aquatic toxicity data for nanomaterials. Additional genotoxicity data were also separately identified during the preparation of a review of the genotoxicity of metal oxides. Provisional datasets summarising the data within these additional articles have been prepared. The most appropriate means of integrating all of these additional data into the final NanoPUZZLES datasets were discussed during the first reporting period and a solution will be developed in the next reporting period.

Two recent actions were taken to help refine the current proposals regarding minimum information standards, which would inform refinement of the final NanoPUZZLES datasets prepared as ISA-TAB-Nano files and made available via Task 1.5. Firstly, work on an exercise to share database/dataset fields amongst participants in the NanoSafety Cluster Databases Working Group was started by NanoPUZZLES. Secondly, it was agreed to lead work on a “Data Completeness” paper, upon invitation from the US Nanotechnology Working Group.

Task 1.2: Chemical Structure Data Curation

A set of criteria for evaluating the quality of data relating to nanomaterial structure, as recorded using ISA-TAB-Nano files, was developed. The structural data, associated with a given nanomaterial record, to be evaluated according to this proposal, included key physiochemical/structural measurements such as size statistics and chemical composition, which have been proposed as critical for uniquely defining a nanomaterial, as well as corresponding electron microscopy images, SMILES files (for fullerenes) and partial representations, such as unit cell crystallographic information format (CIF) files, for those nanomaterials, such as metal oxides, for which a single molecular species cannot be defined. The proposed criteria to be applied consider this information to be of higher quality if it is more complete, has been recorded using a standardised terminology, is defined using sufficiently precise terms and has undergone quality assurance. Based upon these criteria, a proposal for assigning quality scores for this information was developed. The selection of the structural information to be recorded and the means of assigning a quality score to this information were based upon a careful review of the literature.
This proposal was described in detail in deliverable D1.2. This created a starting point for further discussion and harmonisation of the criteria with other projects.

Task 1.3: Toxicity Data Quality Assignment

Nano-QSAR models for (eco)toxicological effects should be built upon the highest quality data available. This requires a suitable definition of (toxicity) data quality and suitable means of ‘scoring’ (or categorising) the quality of the available toxicity data to allow for prioritisation.
A proposal for assigning a quality ‘score’ to the toxicity data within the final NanoPUZZLES datasets (see Task 1.5) was developed based upon a review of existing approaches to evaluating (toxicity) data quality, developed for conventional chemicals or nanomaterials, as well as key concerns regarding nanotoxicology studies that have been raised in the experimental literature. The key considerations highlighted by the literature review were incorporated into checklists for assignment of a given toxicity measurement, as recorded within the NanoPUZZLES data collection, to categories denoting its “reliability”, “relevance” and “adequacy” as defined in the toxicology literature. The assignment procedure was designed to be as transparent and as objective as possible.

Task 1.4: Data Mining Grouping and Category Formation

Initially, a detailed review of the literature was carried out to ensure that all relevant publications relating to grouping strategies for nanomaterials were considered before developing approaches within NanoPUZZLES. The conclusion drawn from this body of literature was that expert-based approaches to grouping of nanomaterials were challenging due to a lack of complete understanding of the features responsible for the dominant nanomaterial toxicity mechanisms (e.g. whether morphology or chemical composition can be considered the dominant feature for a given nanomaterial). Indeed, this conclusion was in keeping with the expectation that pattern recognition approaches would need to be investigated as proposed in the NanoPUZZLES Description of Work. Hence, an additional literature review was carried out to identify suitable pattern recognition approaches to be investigated for category formation.

Pattern recognition methodologies which were considered interesting to explore, for the purpose of category formation, were as follows: (i) calculating the Euclidean distance between nanomaterial samples based on the descriptor most correlated to toxicity, followed by hierarchical clustering according to Ward’s method and Sneath’s stopping criterion; (ii) calculating a non-linear distance between nanomaterial samples, using the Random Forest algorithm, followed by clustering according to the recently published PFClust clustering algorithm. A detailed description of this work was presented in deliverable D1.4.
At this stage in the project (summer of 2014), problems were identified with the provisional NanoPUZZLES ISA-TAB-Nano data collection (developed within the first reporting period), using the validation software, which was available from the MODERN project. It was resolved to investigate these issues via arranging discussions with the MODERN project team (see the description of work carried out towards Task 1.5 during the second reporting period) and focus analysis on other datasets which were available to the NanoPUZZLES project.

Two datasets were identified as potentially useful for Task 1.4: (a) Excel exports of the Nanomaterial Biological Interactions (NBI) Knowledgebase datasets1, as obtained by the NanoPUZZLES team in the first reporting period; (b) the dataset reported by Pathakoti et al. 2 Dataset (a) was of interest due to the large size (a total of 120 nanomaterials were available) and diverse range of nanomaterial types associated with biological testing on embryonic zebrafish using a consistent experimental protocol. Dataset (b) was of interest due to the experimental assignment, by the authors of the original study, of different mechanisms for cytotoxicity to different nanomaterials in the dataset. These dataset characteristics were all relevant for the assignment of different types of nanomaterials to groups associated with a common mechanism for a given endpoint i.e. mechanistically justified category formation.

The results obtained suggested the investigated approaches might be of some value for making “read-across” predictions. However, the categories obtained via these pattern recognition techniques (i.e. the clusters) had no relationship to the experimentally assigned mechanisms proposed for the Pathakoti et al. dataset and did not clearly correspond to specific mechanisms for the NBI dataset. Hence, the scientific validity of the categories obtained was uncertain. This emphasised the importance of continued investigation of grouping approaches in WP4.

Task 1.5: Development of a Data Set including the Inventory of NPs

The relevant resources developed within the first reporting period were significantly extended to yield the final NanoPUZZLES data collection, which was publicly released via various online, searchable databases at the end of December 2015. The publicly released data resources are as follows: a set of ISA-TAB-Nano datasets, containing (meta)data extracted from prioritised nanotoxicology literature references, as well as a list of evaluated references used to identify sources of data to be used within the project.

A detailed description of most of this work was presented in deliverable report D1.5. Hence, a brief summary of the work reported in that deliverable report and carried out during the second reporting period is provided here.

The initial ISA-TAB-Nano data collection created within the first reporting period was significantly expanded to include data for an additional 140 (nominal) nanomaterials, extracted from six additional references. To support this, significant work was carried out to expand the annotated list of articles from which nanosafety relevant data could be extracted into the ISA-TAB-Nano data collection and/or be used to support the modelling activities of WP4. This entailed additional searches for references providing data relevant to the NanoPUZZLES prioritised endpoints (cytoxicity, genotoxicity and zebrafish toxicity), including incorporation of work carried out in the second reporting period towards Golbamaki et al.3 as well as additional information related to the availability of physicochemical characterisation or otherwise related the quality of the corresponding nanotoxicology results identified during work towards Task 1.3 in the first reporting period and subsequent discussions with other projects during the second reporting period e.g. whether assay interference was accounted for.

The provisional scoring schemes developed in the first reporting period and reported in deliverables D1.2 and D1.3 were reviewed, both in terms of the suitability and scientific validity of the approaches as well as to ensure that they could be applied in a time-efficient manner to the ISA-TAB-Nano datasets. This review process included approaching an experimentalist involved in the MODENA-COST Action, as well as discussions with researchers in the MOD-ENP-TOX project, and other participants in the NanoSafety Cluster Databases Working Group and U.S. Nanotechnology Working Group within the context of developing an article, for which work was led by NanoPUZZLES, addressing the challenges associated with evaluating the quality and completeness of curated nanomaterial data.4 This review process led to the development of a more pragmatic scoring scheme, which was used to assign completeness/quality scores to the publicly released ISA-TAB-Nano datasets (Please note that the quality scores can be retrieved via the “README” file distributed within each publicly released ISA-TAB-Nano dataset).

In addition to simply expanding the ISA-TAB-Nano datasets developed within the first reporting period, the tools developed to support the creation of these datasets needed to be updated for two reasons:
(1) to enable compatibility with the nanoDMS database developed within the MODERN project;5
(2) to support the integration of additional kinds of (meta)data, such as in vivo cytotoxicity and zebrafish mortality data, identified as being important but which were not captured by earlier iterations of the NanoPUZZLES ISA-TAB-Nano templates. Hence, the Excel templates6 and Python program7 (for generating tab-delimited text versions of the datasets which could be integrated within nanoinformatics resources such as the MODERN nanoDMS database) were updated and have also been publicly released. In order to address issue (1), errors in all ISA-TAB-Nano datasets – including those developed within the first reporting period – needed to be manually fixed. Detailed discussions with the MODERN project team were required in order to address issue (1) and the NanoPUZZLES project team was also able to provide useful feedback to the MODERN project team, which they incorporated into refining their software.

A detailed explanation of the resources developed within NanoPUZZLES for creating ISA-TAB-Nano datasets is provided in Marchese Robinson et al.8 a joint NanoPUZZLES /MODERN article, which was written during the second reporting period. In the course of writing this article, valuable feedback was obtained from members of the MODERN project, as discussed above, as well as the eNanoMapper project and U.S. Nanotechnology Working Group, which helped to improve the article and make it a valuable reference for nanosafety researchers aiming to make use of ISA-TAB-Nano in the future. It should be noted that this article refers to a snapshot of these resources, which have subsequently been extended and the extended versions publicly released.

An overview of all ISA-TAB-Nano datasets planned for public released, along with the spreadsheet of evaluated references, was provided in deliverable report D1.5. Subsequent to submitting deliverable report D1.5 further improvements were carried out to the relevant data resources prior to their being publicly released. These improvements entailed addressing a list of possible errors and omissions associated with curation, which were identified during the preparation of these datasets, as well as further analysis of possible systematic errors in the datasets – such as ISA-TAB-Nano Material file names, which were inconsistent with their “Material Source Name” identifiers.

The final ISA-TAB-Nano datasets correspond to 240 Material files, 20 Study files and 74 Assay files. (However, in some cases, the Material files correspond to nominal nanomaterials, with the curated size measurements outside the 1 – 100 nm range, and data were not available for all endpoints for all materials.) Toxicity data collected were primarily for different kinds of cytotoxicity assays, with some (embryonic) zebrafish mortality and genotoxicity data also collected. In addition, corresponding physicochemical data were collected for a range of endpoints, such as size, zeta potential and dissolution.

The final spreadsheet of evaluated references, from which prioritised references were identified to support the modelling work within NanoPUZZLES and/or to develop the ISA-TAB-Nano datasets, corresponds to a set of 336 primary experimental references annotated with various kinds of metadata, which were used to select prioritised references. These metadata include the number of nanomaterials studied (or, in some cases where data were not available for all nanomaterials, the number of nanomaterials for which at least some data were available) and the availability of different kinds of data for various nanosafety related endpoints.

In order to identify possible means for making all publicly released datasets available via a searchable database and to ensure that they would be available for future nanosafety researchers, a review of various options was considered. It was decided that, in addition to submitting the ISA-TAB-Nano datasets to the nanoDMS system developed within the MODERN project – the Zenodo and FigShare online resources were appropriate to serve as long term, searchable hosting options for all publicly released dataset files. These resources, as well as the tools developed to facilitate the creation of ISA-TAB-Nano datasets, are publicised via the NanoPUZZLES website,9 although this was not felt to be the most suitable, long-term solution for hosting the datasets themselves.

The final versions of all ISA-TAB-Nano datasets10 and the list of evaluated references11 were made available from both Zenodo and FigShare12 at the end of December 2015. In addition, all cytotoxicity and genotoxicity ISA-TAB-Nano datasets were validated and made available via the searchable, ISA-TAB-Nano based, online database (nanoDMS) developed within the MODERN project.13 All datasets are made available via the Creative Commons Attribution License, which allows them to be freely redistributed and reused as long as the original dataset records are cited.

WP2: NanoDESC:

Task 2.1: Evaluation of the existing systems for structural characterization of NPs

We provided information about the descriptors, which are used for QSAR modelling of engineered nanoparticles. First, we evaluated the classical descriptors, used for classical QSAR and then we discussed the appropriate descriptors used for nanoparticles. However, there are specific parameters which go beyond the traditional approach(es) for QSAR and may be profitably used as descriptors for nanoparticles.

The emphasis of Task 2.1 was to evaluate existing strategies, as reported in published studies, and review the possible ways to proceed, identifying cases which appear more consolidated, characterising the premises which make these strategies applicable.
We compared the descriptors for nanoparticles with those used for bulk materials. Moreover, we introduced case studies of new QSAR models developed by using various descriptors within NanoPUZZLES project.

Furthermore, we identified the use of traditional and new descriptors, as applied within a series of examples. The use of these descriptors has been discussed in detail, defining the domain and limitations of their application. An appropriate review of the descriptors was presented in deliverable D2.1.

Task 2.2: Development of simplified molecular models sufficient to characterize the whole structure

The objective of Task 2.2 is to develop simplified molecular models to characterise the whole structure of nanoparticles. There are several formats to record the chemical structure, such as SMILES-based, orbital graphs and InChI-based methods.

The description and characterisation of nanoparticles by traditional schemes is limited, because
(i) the molecular architecture of nanoparticles is extremely large and extremely complex;
(ii) there are specific interactions between different parts of the nanosystems which cannot be represented topologically and/or by means of molecular mechanics and quantum mechanics. The modeller may combine descriptors and formats from different backgrounds where the information is not redundant.

A possible way to organise a model for endpoints related to nanomaterials, at the present time, is to assume that the measured endpoint is a mathematical function of all available eclectic information. This means that, in order to define a predictive model for an endpoint related to nanomaterials, the traditional paradigm ‘Endpoint = F(molecular structure)’should be replaced with ‘Endpoint = F(eclectic information)’.
The eclectic information can be
(i) atomic composition;
(ii) conditions of synthesis/preparation of the nanomaterial;
(iii) the features of nanomaterials related to their manufacturing. This list can be easily extended (e.g. size, porosity, symmetry, electromechanical properties, etc.). These additional descriptors have to be obtained from experimental measurements, associated to the specific nanomaterial, and cannot be calculated.
Many other factors able to influence the numerical values of the measured endpoint, and thus the model can be detailed according to new available information (irradiation, dark, temperature, etc.) related to the specific phenomenon under investigation.

The statistical and operative schemes of the QSAR models should be adopted as usual, including a broader definition of descriptors in their applicability domain. This means that if we extend the definition of the descriptors to other non-classical features, these have also to be used for the characterisation of the applicability domain of the model.

Two papers have been published (see below) in which we used the proposed format for modelling nanomaterials. These studies have been done by IRFMN and UG.
In addition, Partner NHRF contributed with QM calculations. NHRF investigated various descriptors in order to be used as a basis to build up predictive models for a nanomaterials endpoint. The work of A. Avramopoulos and G. Leonis has been continued and extended in their contribution to WP4.
Toropova A.P. Toropov A.A. Puzyn T., Benfenati E., Leszczynska D., Leszczynski J. (2013): Optimal descriptor as a translator of eclectic information into the prediction of thermal conductivity of micro-electro-mechanical systems. Journal of Mathematical Chemistry 51, 2230-2237.
We specify that the work done by the US authors have been financed within NanoBRIDGES, while the work done by European authors have been financed by NanoPUZZLES. The work done within NanoPUZZLES refers to the calculation of the descriptors and preparation of the QSAR model.
Toropov A.A. Toropova A.P. Puzyn T., Benfenati E., Gini G., Leszczynska D., Leszczynski J. (2013): QSAR as a random event: modeling of nanoparticles uptake in PaCa2 cancer cells. Chemosphere 92, 31-37.

Task 2.3 Development of descriptors for the nanostructure (“nano-descriptors”) of five types

The most promising categories of descriptors potentially useful in developing new predictive models for the activities and properties of the nanoparticles as well as examples for their applicability and demonstration of the utility of this kind of descriptors are listed below:

Topological descriptors

Topological descriptors can be calculated with a series of tools, such as molecular graphs, SMILES and InChi. They can also be based on physicochemical parameters. As an application and demonstration of the utility of this kind of descriptors, we studied: (1) cytotoxicity of metal oxide nanoparticles, (2) mutagenicity of fullerenes and multi-walled carbon-nanotubes, (3) membrane damage by TiO2 and ZnO nanoparticles, as well as (4) cell viability of human embryonic kidney cells exposed to SiO2 nanoparticles. This study has been done by IRFMN:
Toropova A.P. Toropov A.A. Rallo R., Leszczynska D., Leszczynski J. (2015): Optimal descriptor as a translator of eclectic data into prediction of cytotoxicity for metal oxide nanoparticles under different conditions. Ecotoxicol. Environ. Safe. 112, 39-45.
Toropov A.A. Toropova A.P. (2014): Optimal descriptor as a translator of eclectic data into endpoint prediction: Mutagenicity of fullerene as a mathematical function of conditions. Chemosphere. 104, 262-264.
Toropov A.A. Toropova A.P. (2015): Quasi-QSAR for mutagenic potential of multi-walled carbon-nanotubes. Chemosphere. 124, 40–46.
Toropova A.P. Toropov A.A. Benfenati E., Puzyn T., Leszczynska D., Leszczynski J. (2014): Optimal descriptor as a translator of eclectic information into the prediction of membrane damage: The case of a group of ZnO and TiO2 nanoparticles. Ecotoxicol. Environ. Safe. 108, 203-209.
Manganelli S., Leone C., Toropov A.A. Toropova A.P. Benfenati E. (2016): QSAR model for predicting cell viability of human embryonic kidney cells exposed to SiO2 nanoparticles. Chemosphere. 144, 995-1001.

Nano-descriptors derived from quantum-mechanical calculations

Based on simply computable, interpretable and reproducible molecular descriptors calculated with Gaussian09, MOPAC2012 and Dragon software we have demonstrated the utility of the quantum-mechanical descriptors in developing new predictive Nano-QSAR models. Quantum-mechanical calculations have been performed at the semi-empirical level with PM6 and PM7 methods and DFT level of theory using B3LYP/6-31++G** and M06-2X/6-31++G**. Adequate and robust regression based Nano-QSAR models for predicting cytotoxicity and genotoxicity of metal oxide nanoparticles as well as cellular uptake of superparamagnetic iron oxide nanoparticles (SPIONs) in human umbilical vein endothelial cells (HUVEC) and pancreatic carcinoma cell line (PaCa2) were established:
Sizochenko N., Rasulev B., Gajewicz A., Kuzmin V.E. Puzyn T., Leszczynski J. (2014): From basic physics to mechanisms of toxicity: Liquid Drop approach applied to develop predictive classification models for toxicity of metal oxide nanoparticles. Nanoscale. 6, 13986-13993.
Golbamaki N., Rasulev B., Cassano A., Marchese Robinson R.L. Benfenati E., Leszczynski J., Cronin M.T.D. (2015): Genotoxicity of metal oxide nanomaterials: review of recent data and discussion of possible mechanisms. Nanoscale. 7, 2154-2198.
Gromelski M., Lewandowska W., Gajewicz A., Puzyn T. (2016): Comparative nano-QSAR modelling of cellular uptake for coated nanoparticles in pancreatic carcinoma cell line (PaCa2) based on calculations at the DFT and semi-empirical level of theory. Manuscript is being prepared.
Lewandowska W., Gromelski M., Gajewicz A., Puzyn T. (2016): Towards understanding mechanisms of cellular uptake of superparamagnetic iron oxide nanoparticles (SPIONs) in human umbilical vein endothelial cells - local versus global nano-QSAR modelling. Manuscript is being prepared.

Descriptors derived from computational processing of microscopic (SEM/TEM/AFM) images

In collaboration with the NanoBRIDGES project, SEM/TEM images of nanoparticles were processed and analysed by using the ImageJ software to obtain new descriptors. We specify that the work done by the US authors have been financed within NanoBRIDGES, while the contribution done by European authors (Partner UG) has been financed by NanoPUZZLES. The work done within NanoPUZZLES refers to the calculation of microscopic images-based descriptors using the algorithm developed within NanoBRIDGES project. Morphological (large scale) features in relation to shape, and size were analyzed. They were expressed in the form of ten key descriptors: area, perimeter, major axis, minor axis, aspect ratio, Maximum Feret's diameter, Minimum Feret's diameter, roundness, circularity, and solidity.
Odziomek K., Ushizima D., Oberbek P., Kurzydłowski K. J., Puzyn T., Haranczyk M. (2016): Scanning electron microscopy image representativeness: morphological data on nanoparticles. Submitted (under review).

In addition, a list of about 260 papers/reports was evaluated to verify the availability of these images as well as cytotoxicity/genotoxicity data, in collaboration with activities within WP1 by LJMU.

Descriptors based on “anisotropy dimensions”

On the basis of the evaluation of the existing literature, the most promising perspective for the near future is to assess if this feature may explain some of the unusual behaviour of the nanomaterials. In other words, this information can be applied a posteriori in case of outliers of the model, to assess if the outlier can be affected by the anisotropic properties, which have been reported for such a material. IRFMN evaluated the feasibility of these descriptors for nanoparticles QSAR modelling.

Descriptors based on graph or fractal theory

Partner IRFMN evaluated the studies already done on fractal dimensions of the nanoparticles in the literature. There are a number of case studies in which fractal dimensions are used as descriptors of nanoparticles to characterize them.
Besides the five kinds of descriptors that we discussed above, which have been anticipated in the DoW, we extended this Task to three other kinds of descriptors, as below:

«Liquid Drop» Model (LDM) descriptors

The physical model of "liquid drop" was used to describe the geometric and volume features of studied metal oxide nanoparticles. In this model a nanoparticle is represented as a spherical drop, where elementary particles are densely packed and the nanoparticle’s density is equal to the density of bulk. A study was performed under the NanoBRIDGES project, which is in collaboration with NanoPUZZLES. We specify that the work done by the US and Ukrainian authors have been funded within NanoBRIDGES, while the contribution done by European authors has been funded by NanoPUZZLES. The contribution done within NanoPUZZLES refers to the development of “liquid drop” model derived descriptors, whereas the final nano-QSAR model has been developed within NanoBRIDGES project. The following paper has been submitted related to calculating nano-descriptors based on the LDM approach. This study has been done by UG.
Sizochenko N., Rasulev B., Gajewicz A., Kuzmin V.E. Puzyn T., Leszczynski J. (2014): From basic physics to mechanisms of toxicity: Liquid Drop approach applied to develop predictive classification models for toxicity of metal oxide nanoparticles. Nanoscale. 6, 13986-13993.
Sizochenko N., Jagiello K., Leszczynski J., Puzyn, T. (2015): How the “Liquid Drop” approach could be efficiently applied for Quantitative Structure–Property Relationship modeling of nanofluids. J. Phys. Chem. C, 119, 25542–25547.

Zero-dimensional (0D) - Constitutional descriptors

Metal electronegativity (χ), the charge of the metal cation corresponding to a given oxide (χox), atomic number and valence electron number of the metal have been used as simple molecular periodic table-based descriptors to build up QSTR models. This model has been described in the following contribution by UG:
Kar S., Gajewicz A., Puzyn T., Roy K. (2014): Periodic table-based descriptors to encode cytotoxicity profile of metal oxide nanoparticles: A mechanistic QSTR approach. Ecotoxicol. Environ. Safe. 107C, 162-169.

Simplex Representation of Molecular Structure (SiRMS) as nanoparticles descriptors

SiRMS method was applied to calculate nano-structural descriptors for studying toxicity of metal oxides nanoparticles to bacteria E. coli. In the framework of this approach, any molecule/particle is represented as a system of different simplexes (tetratomic fragments of fixed composition).
Considering that the objects of research were metal oxides, we used shorter fragments of molecules (for 1 to 3). The connectivity of atoms in a simplex, atom type, and type of chemical bond have been considered. Particular atoms in a simplex are differentiated on the basis of atomic refraction, electronegativity and van-der-Waals interactions. In this study, all atoms were divided into groups corresponding to their electronegativity (A<1.3 Sizochenko N., Rasulev B., Gajewicz A., Kuzmin V.E. Puzyn T., Leszczynski J. (2014): From basic physics to mechanisms of toxicity: Liquid Drop approach applied to develop predictive classification models for toxicity of metal oxide nanoparticles. Nanoscale. 6, 13986-13993.
Sizochenko N., Jagiello K., Leszczynski J., Puzyn, T. (2015): How the “Liquid Drop” approach could be efficiently applied for Quantitative Structure–Property Relationship modeling of nanofluids. J. Phys. Chem. C, 119, 25542–25547.

Task 2.4: Development of algorithms for calculating nanodescriptors

The objective of Task 2.4 was to implement a system (database and software) for the free and simplified use of the nanodescriptors on the Internet. This system includes an updated version of the CORAL software, developed within the EC project Marie Curie Fellowship (n. MIF1-CT-2006-039036, CHEMPREDICT), which can develop SMILES based descriptors and QSAR models.
IRFMN improved and adapted the original software, developed for QSAR models based on SMILES structures, to the use of nanomaterials. In this case, CORAL accepts a number of other features, not related to the chemical structure, which derive from values to be inserted by the user, such as physicochemical properties, experimental conditions etc.

The CORAL software can be freely downloaded on the CORAL website, at http://www.insilico.eu/coral/SOFTWARECORAL.html. According to the statements of Task 2.4 the software has been made freely available also on the NanoPUZZLES website, at http://www.nanopuzzles.eu/results-products.
In order to build up a model for predicting nanomaterials toxicity using CORAL, the representation and characterization of nanomaterials should be organized according to the following principles:
The endpoint is any physicochemical or biochemical representative parameter, which is available for a group of similar nanomaterials;
The paradigm for building up a model is the following:

Endpoint = F (Eclectic Information)

The eclectic information can include atomic composition, conditions of synthesis, features of nanomaterials related to their manufacturing, as well as size, porosity, electromechanical properties, etc.
This information needs to be encoded in the quasi-SMILES according to three main steps:
Use of an alphabetic character/ numeric symbol to encode each selected feature. For example if the size and concentration are considered as parameters for the characterization, they will be indicated as “A” and “B” respectively.
Normalization of the considered features (optional step).
Discrimination of each feature into one of n categories, according to its experimental values.

The so-called ‘optimal descriptors’ for nanomaterials are calculated with the correlation weights for these encoded attributes, CW(Codek), as the following:
DCW (Treshold, Nepoch) = ∑ CW(Codek)
The numerical values of CW are function of Threshold (T) and number of epochs (Nepoch), which are parameters of the Monte Carlo optimization. The threshold is a tool to define two classes of features: rare (noise) and not rare, i.e. active. The optimal descriptors are calculated with the correlation weights of active features. Correlation weights for rare features are equal to zero; these are not involved in building up model.
The endpoint is dependent on the optimal descriptors:

Endpoint= C0 + C1 x DCW (T, Nepoch)
The models for nanomaterials should be built and validated on an internal set (split into sub-training, calibration, and test sets), and then on an external validation set.
The advantages are that in this way any feature can be easily introduced within the modeling scheme. However, this is immediately related to a constraint, and thus a possible limitation: indeed, if the model requires as component a feature, which is an experimental value, this value has to be provided for the prediction of the property of the nanomaterial. Of course, this limitation is quite a general one, which often applies to most models for nanomaterial.
An appropriate tutorial for the use of the CORAL software to develop free available nanoQSAR models was presented in deliverable D2.4.
Many models for nanomaterials have been developed within the CORAL software within Task 2.4 to predict different endpoints, such as nanoparticles uptake in PaCa2 cancer cells, thermal conductivity of Micro-Electro-Mechanical Systems, membrane damage by TiO2 and ZnO nanoparticles, mutagenicity of fullerene and cytotoxicity by metal oxides. In particular, as an example, some of them has been made freely available within CORAL and can be used for prediction and as a template for other modeling purposes. The full list of publications produced under NanoPUZZLES has been made available on the CORAL website at http://www.insilico.eu/coral/ABOUT_US.html.

The publications produced within Task 2.4 in collaboration with WP3 are listed below:
Worachartcheewan, A., Mandi, P., Prachayasittikul, V., Toropova, A.P. Toropov, A.A. Nantasenamat, C. (2014): Large-scale QSAR study of aromatase inhibitors using SMILES-based descriptors. Chemometrics and Intelligent Laboratory Systems, 138, 120-126.
Duchowicz, P.R. Fioressi, S.E. Bacelo, D.E. Saavedra, L.M. Toropova, A.P. Toropov, A.A. (2015): QSPR Studies on Refractive Indices of Structurally Heterogeneous Polymers. Chemometrics and Intelligent Laboratory Systems, 140, 86–91.
Veselinović, J. B., Toropov, A. A., Toropova, A.P. Nikolić, G.M. Veselinović, A.M. (2015): Monte Carlo Method-Based QSAR Modeling of Penicillins Binding to Human Serum Proteins. Archiv der Pharmazie, 348, 62-67.
Toropova, A.P. Toropov, A.A. Rallo, R., Leszczynska, D., Leszczynski, J. (2015): Optimal descriptor as a translator of eclectic data into prediction of cytotoxicity for metal oxide nanoparticles under different conditions. Ecotoxicology and Environmental Safety, 112, 39–45.
Toropov, A.A. Toropova, A.P. (2015): Quasi-QSAR for mutagenic potential of multi-walled carbon-nanotubes. Chemosphere, 124, 40–46.
Toropova, A.P. Toropov, A.A. (2015): Mutagenicity: QSAR - quasi-QSAR - nano-QSAR. Mini-Reviews in Medicinal Chemistry, 15, 608-621.
Toropova, A.P. Toropov, A.A. Veselinović, J. B., Veselinović, A.M. Benfenati, E., Leszczynska, D., Leszczynski, J. (2015): Application of the Monte Carlo method to prediction of dispersibility of graphene in various solvents. International Journal of Environmental Research, 9, 1211-1216.
Toropov, A.A. Toropova, A.P. Veselinović, A.M. Veselinović, J. B., Nesměrák, K., Raska Jr, I., Duchowicz, P. R., Castro, E. A., Kudyshkin, V.O. Leszczynska, D., Leszczynski, J. (2015): The Monte Carlo method based on eclectic data as an efficient tool for predictions of endpoints for nanomaterials – two examples of application. Combinatorial Chemistry & High Throughput Screening, 18, 376-386.
Toropova, A.P. Toropov, A.A. (2015): Quasi-SMILES and nano-QFAR: United model for mutagenicity of fullerene and MWCNT under different conditions. Chemosphere, 139, 18-22
Toropova, A.P. Toropov, A.A. Kudyshkin, V.O. Rallo, R. (2015): Prediction of the Q-e parameters from structures of transfer chain agents, Journal of Polymer Research, 22, 128.
Manganelli, S., Leone, C., Toropov, A.A. Toropova, A.P. Benfenati, E. (2015): QSAR model for cytotoxicity of silica nanoparticles on human embryonic kidney cells. Materials Today: Proceedings (in press).
Manganelli, S., Leone, C., Toropov, A.A. Toropova, A.P. Benfenati, E. (2016): QSAR model for predicting cell viability of human embryonic kidney cells exposed to SiO2 nanoparticles. Chemosphere, 144, 995-1001.

WP3: NanoINTER:

Task 3.1: Development of a protocol for the study of large interacting systems

For the development of the protocol to study large interacting systems we have performed extensive calculations, which have involved:
The use of a large array of techniques, (e.g. semi-empirical-PM6-, HF, DFT- B97D, wB97XD, M062X, B3LYP, PBE0-, MP2, molecular dynamics) and basis sets (e.g. 6-31G*, 6-31+G*, 6-311G*). We have also investigated the performance of several ab initio techniques for the calculation of the intermolecular interaction energy (e.g. the Su & Li approach, the Kitaura and Morokuma method, the effective fragment potential approach, a variational perturbation scheme) as well as the Atoms-In-Molecules technique for the computation of the hydrogen bond (HB) energy.
The efficient use of geometry optimisation techniques (e.g. B97D/6-31G*, PM6).
The computation of several properties (e.g. the interaction energy, the energy of HOMO and LUMO, ionisation potential, electron affinity, dipole moment, polarisabilities and first hyperpolarisabilities).
Testing the calculations on a variety of nanoparticles (NPs), for example: C24H12, C84H20, C24H12, C114H30, C222H42, C366H54; C60, (TiO2)n, SWCNT(C360). These have been used to compute their structure and interaction properties.
Testing the calculations on a variety of biological molecules (BMs). For example: proteins or fragments of them (e.g. of HIV-1PR, renin, GPCR, human serum albumin). The set of NPs and BMs considered has been employed to compute their structure and interaction properties.

The above experimentation led to the proposed computational protocol, which involves:
A DFT technique for computing the interactions of NP-NP systems (e.g. graphene/C60, graphene/(TiO2)n, SWCNT/C60). Recommendation: B97D/6-31G*.
The most efficient ab initio techniques for calculating the interaction properties of BM-BM systems (e.g. Gly49-Ile50-Gly51–water–Gly49´-Ile50´-Gly51´, L10I-L63P-A71V(SQV, Arg8–SQV). Recommendation: The AIM method for the computation of the HB energy and the Su and Li method for the calculation of the interaction energy.
Molecular dynamics for studying large BM-NP (e.g. human serum albumin-C60) or BM-BM (e.g. JDTic-GPCR) systems. Recommendation: MM-PBSA for the calculation of the interaction energy and the AIM method for the computation of the HB energy.

Task 3.2: Implementation of techniques for the resolution of the interaction energy into various contributions

The interaction energy of a series of fullerene derivatives with human serum albumin (HSA), or appropriately defined models of HSA, has been computed and resolved into a variety of meaningful contributions by employing ab initio methods (e.g. Su and Li, Kitaura and Morokuma, the effective fragment potential approach, the fragment molecular orbital method). Molecular dynamics for resolution of the binding free energy (MM‒PBSA) has also been employed. Considerable care has been taken in the validation of the methods employed. After a detailed analysis of the large body of the computed data, we recommend, for the resolution of the interaction energy:
For small or average size systems the use of Su and Li at the DFT level of theory (e.g. B3LYP/3-21G*) provides valuable information on the nature of the interaction mechanism;
The fragment molecular orbital method provides very useful information for the whole interacting system, but it is computationally demanding; and
For large interacting systems the employment of MM‒PBSA.

Task 3.3: Development of techniques for the study of the environment (e.g. solvent) on the interacting system

The objective of this deliverable was to develop of techniques for the study of the environment (e.g. solvent) on the interacting system. The most important conclusions are as follows:
Interactions of NPs with small molecules: (i) ΔΕ (solubilization energy) depends on ε as well as the functional groups bonded to C60 core, and (ii) increase of ε leads to an increase of |ΔΕ|.
Validation of the employed method: the model used for the computation of the solubility (S) is in satisfactory agreement with the experimental data.
Complex aqueous solutions involving two kinds of NPs: The solvent may have a significant effect on some properties (e.g. the average polarizability) of the interacting system.
Reactivity studies of fullerene derivatives: the solvent has a negligible effect on the reactivity of the considered [60]fullerene derivatives.
Quantum mechanical study of aggregation: it has been found that aggregation (in the gas phase or in solution) has a small effect on the considered properties of (C60)n.
Molecular dynamics study of aggregation: molecular aggregation of [60]fullerene systems is obtained spontaneously in aqueous solution, as represented by stable hydrophobic clusters throughout the MD simulations. On the other hand, water-soluble fullerene structures do not form agglomerates in water regardless their concentration and initial member separation.

The proposed set of techniques for the study of the effect of the environment on the interacting systems involves: (i) DFT techniques for the study of small interacting systems; (ii) semi-empirical approaches (e.g. PM6) for the calculation of the properties of larger interacting systems; (iii) molecular dynamics (MD) for the study of large interacting systems. In addition, MD provides valuable information for the dynamic development of the interaction system with time; (iv) a relationship is proposed for the calculation of the solubility of fullerenes. The solubilities produced by this relationship are in satisfactory agreement with the experiment. In addition, this relationship provides a framework for the understanding of the solubilization and the solute-solvent interactions.

Task 3.4: Design/recognition of functional groups, which seriously reduce the genotoxicity and increase the solubility of the considered NPs

The objective of this deliverable was to connect the genotoxicity/mutagenicity of a series of fullerenes, C60Fn, where n=10, 12, 14, 16 & 18 with their solubility. We have found from the literature that EHOMO increases with mutagenicity. This information allowed to connect the mutagenicity with the solubility of C60Fn. It was shown that an increase in n was followed by an increase in solubility and a decrease in mutagenicity. We have also demonstrated that mutagenicity increases with lipophilicity (logP), QF (the sum of the charges on the fluorine atoms of C60Fn), E(S-T), the dipole moment, the average polarizability, and the first hyperpolarizability. Therefore, D3.4 presented evidence for a relationship connecting mutagenicity with solubility and a set of important electronic descriptors. The solubility data of C60Fn have been extracted from the literature. The electronic descriptors have been computed by employing the B3LYP/6-31*G method, while logP was estimated using the Schrödinger software. We have also studied the interaction of C60Fn with DNA and human DNA topoisomerase II-alpha (HT2a) by performing molecular dynamics and MM–PBSA free energy calculations. Significant interactions have been found between C60Fn and both systems, particularly between C60Fn and HT2a. These interactions may induce an undesirable effect on the DNA function.

A detailed description of this work has been presented in deliverable D3.4 as well as in the following manuscript: Papavasileiou K., Avramopoulos A., Leonis G., Papadopoulos M. (2016): Computational Investigation of Fullerene DNA Interactions: Implications of Fullerene’s Size, and Functionalization on DNA Structure and Binding Energetics (manuscript submitted to Journal of Chemical Theory and Computation - under review).

Task 3.5: Study of factors affecting the interaction of the selected systems, NP/{biological molecule}

The objective of deliverable D3.5 was the study of factors that affect the interactions between nanoparticles (NPs) and biological molecules (BMs). For this purpose, we have used molecular dynamics (MD) techniques and ab initio calculations with different functionals (B97D, M062X) and basis sets (6-31G*, 6-31G**, 6-31++G**) in order to study how the: i) chemical composition, ii) size and shape, iii) aggregation, iv) surface charge, and v) contamination of NPs affect a series of properties, such as the structure, the binding free energy (ΔEbind), the interaction energy (Eint), the energy of the highest occupied molecular orbital (EHOMO), and the energy of the lowest occupied molecular orbital (ELUMO) of several NP-BM systems.

As NPs we have used: a) molecular graphene (C84H24, C60B12N12H24, C78H24FeN4), b) C60 and (C60)4, c) two single wall carbon nanotubes (C360H40, C180H20), d) a series of functionalized C60 derivatives, and e) C60Fn, n=10, 12, 14, 16, & 18; compounds (a-c) and (d-e) were employed for the ab initio and MD calculations, respectively. For BMs, the following systems were used: a) five nucleobases (NBs), thymine (T), cytosine (C), guanine (G), adenine (A) and uracil (U), b) the guanine tetramer (G4), c) human serum albumin (HSA), d) human DNA topoisomerase II-alpha (HT2a), and e) a DNA sequence; systems (a-b) and (c-e) were employed for the ab initio and MD calculations, respectively. The ab initio study on the effect of size and shape of NP was performed on complexes: C84H24-X/Y, C60-X , C360H40-X/Y, C180H20-X, where X=Guanine and Y=Adenine.

A detailed description of most of this work was presented in deliverable report D3.5. Hence, the main findings of this deliverable are summarized as follows:

Chemical composition: Regarding fullerene–HSA complexes, the apo form of albumin is more flexible and less compact than the fullerene-bound forms of the protein, while IIIA complexes are more mobile than IIA complexes. Fullerene binding to IIA is potentially associated with allosteric modulation (secondary structure changes of IIIA and heme binding sites). All fullerene analogs were practically stable in the binding sites of HSA; compounds were stabilized in either cavity of HSA through diverse HB interactions that involve residues Glu188/292, Lys199, His242, Arg218/257 (IIA site), and Arg410, Glu382–Asn391, Lys413/414, Phe488–Glu492 (IIIA site). All fullerene analogs display significant binding affinities for HSA, with free energy calculations ranging from –10 to –70 kcal/mol. IIA binding is approximately 10-fold more efficient that IIIA binding. Residues Glu292–Asn295, and Cys437/448 contribute favorably to IIA binding, while Glu382/383/492, Leu387, Tyr411 and Val415 were associated with effective binding to IIIA site. Binding is driven by van der Waals and nonpolar contributions, while electrostatics disfavor fullerene–HSA complex formation. Long and negatively charged groups attached to the fullerene core are necessary to enhance interactions between ligands and HSA, and thus resulting in effective binding. We also showed that it is possible to bring positively charged species in the IIA cavity, using fullerenes as carriers, since binding was found to be comparable with their negatively charged and polar counterparts. HT2a complexes with fluorinated derivatives had RMSD values, which converged after ~50 ns and the active site of the protein appeared relatively flexible in all systems. C60F12 showed enhanced binding with van der Waals being the major contribution. Similarly, DNA complexes became particularly stable after 100 ns of simulation, however, C60F12 showed significant conformational changes. All complexes had high binding energies and energetic contributions came mostly also from van der Waals and the nonpolar solvation terms.

Size and Shape: In a series of NP-G systems, where NP = C84H24C60, C360H40, C180H20, it was shown that the surface area available for interaction is an important factor. The larger surface curvature results to a larger interaction energy.

Particle Aggregation: Significant structural changes on a guanine tetramer (G4) were observed, by studying the interaction of a (C60)4-G4 complex. It was found that while the isolated G4 structure is planar, the relaxed G4 (optimized in the G4–(C60)4) deviates significantly from planarity. Smaller deviations in distances between C60 molecules of (C60)4-G4 were computed. Both the method and the basis set are important for the computation of ΔEint. By employing the M062X/6-31G** method, significant binding (–18.6 kcal/mol) for (C60)4-G4 is computed.

Surface Charge: The study of the interaction between charged C60 and a guanine molecule revealed significant changes on ΔEint. It was computed that the charged C60 fragment significantly enhances the value of ΔEint compared with the neutral one. A significant change in the orientation of the guanine molecule (in C60–G complex) was observed upon changing the charge of the C60 fragment.

Contamination: The effect of the contamination of NPs on the interaction energy, was examined by studying several NP-BM systems, where NP = C84H24, C60 B12N12H24 and C78H24FeN4 and BM= thymine (T), cytosine (C), guanine (G), adenine (A) and uracil (U). Using a series of methods for the computation of ΔEint, it was found that C84H24 –G and C84H24 –U have the largest and the smallest interaction energies. The adequacy of the methods used for ΔEint was validated through literature data. The same trend, found for ΔEint of C84H24 –BM systems, was observed for the interaction-induced polarizability. This is attributed to the stabilization of the complex due to van der Waals forces, and is proportional to the polarizability of the interacting species. The strength of the interaction energy depends on the type of the contaminant of C84H24. The presence of a metal atom (Fe), gives the highest interaction energy. It was shown that ΔΕint of C78H24FeN4-G is 2.1 times larger than that of C84H24-G, while ΔΕint of C60B12N12H24-G is 1.4 times larger than the corresponding value of C84H24-G. It was found that the type of contaminant (B12N12, FeN4) significantly changes the |HLG|, mostly attributed to the increase of EHOMO.

WP4: NanoQSAR:

Task 4.1: Investigating the impact of size on the phys/chem properties of NPs

A detailed description of this work was presented in deliverable report D4.1. Hence, a brief summary of the work reported in that deliverable report and carried out during the reporting period is provided below. The methodology of investigating the size-dependent electronic properties of metal oxides nanoparticles (regardless of their stoichiometry and symmetry) that might be considered as potential nanodescriptors that are vital for Nano-QSAR and Nano-QSPR studies was identified. In addition, simple relationships for re-calculating structural features (i.e. descriptors) computed at the semi-empirical level to more sophisticated level of theory (i.e. DFT methods with B3LYP and M06 functionals iii) Hartree-Fock methods; and iv) post-Hartree–Fock ab initio methods including Møller–Plesset perturbation theory (MP2)) were established.

Finally based on obtained results, linear regression models were applied to describe the relationships between size (number of atom in cluster) and the electronic properties. Developed and validated methodology would be applied for other nanoclusters to calculate properties that are considered as potential nanodescriptors for Nano-QSAR and Nano-QSPR models.

Task 4.2: Developing Nano-QSAR models of toxicity and environmentally relevant phys/chem endpoints

We demonstrated that by applying Nano-QSAR and Nano-QSPR models the biological activity (e.g. cytotoxicity, mutagenicity, genotoxicity, cellular uptake, membrane damage) and phys/chem properties (e.g. thermal conductivity, photocatalytic activity, viscosity, dispersibility in various solvents) of metal oxides nanoparticles, fullerenes, multi-walled carbon-nanotubes, nanofluids, graphene as well as superparamagnetic iron oxide nanoparticles (SPIONs) decorated with organic molecules could be estimated with a similar level of accuracy as provided by experiments.

A detailed description of this work has been presented in deliverable D4.2 as well as in the following papers, which have been already published, are under review and/or will be submitted until April 2016:
Malankowska A., Mikolajczyk A., Mulkiewicz E., Nowaczyk G., Gajewicz A., Nischk M., Puzyn T., Jurczak P., Hirano S., Jurga S., Zaleska-Medynska A. (2016): Combined experimental and computational approach to develop efficient and environmentally safe photocatalysts based on Au/Pd-TiO2 nanoparticles. ACS Catalysis (under review).
Manganelli S., Leone C., Toropov A.A. Toropova A.P. Benfenati E. (2016): QSAR model for predicting cell viability of human embryonic kidney cells exposed to SiO2 nanoparticles. Chemosphere. 144, 995-1001.

Sizochenko N., Rasulev B., Gajewicz A., Mokshyna E., Kuzmin V.E. Leszczynski J., Puzyn T. (2015): Causal inference methods to assist in mechanistic interpretation of classification nano-SAR models. RSC Advances. 5, 77739-77745.
Sizochenko N., Jagiello K., Leszczynski J., Puzyn, T. (2015): How the “Liquid Drop” approach could be efficiently applied for Quantitative Structure–Property Relationship modeling of nanofluids. J. Phys. Chem. C, 119, 25542–25547.
Toropova A.P. Toropov A.A. Rallo R., Leszczynska D., Leszczynski J. (2015): Optimal descriptor as a translator of eclectic data into prediction of cytotoxicity for metal oxide nanoparticles under different conditions. Ecotoxicol. Environ. Safe. 112, 39-45.

Toropov A.A. Toropova A.P. (2015): Quasi-QSAR for mutagenic potential of multi-walled carbon-nanotubes. Chemosphere. 124, 40–46.

Toropova A.P. Toropov A.A. (2015): Mutagenicity: QSAR - quasi-QSAR - nano-QSAR. Mini-Reviews in Medicinal Chemistry, 15, 608-621.

Toropova A.P. Toropov A.A. Veselinovic J. B., Veselinovic A.M. Benfenati E., Leszczynska D., Leszczynski J. (2015): Application of the Monte Carlo method to prediction of dispersibility of graphene in various solvents. Int. J. Environ. Res. 9, 1211-1216.

Toropov A.A. Toropova A.P. Veselinovic A.M. Veselinovic J. B., Nesmerák K., Raska Jr I., Duchowicz P. R., Castro E. A., Kudyshkin V.O. Leszczynska D., Leszczynski J. (2015): The Monte Carlo method based on eclectic data as an efficient tool for predictions of endpoints for nanomaterials - two examples of application. Com. Chem. High T. Scr. 18, 376-386.

Toropova A.P. Toropov A.A. (2015): Quasi-SMILES and nano-QFAR: United model for mutagenicity of fullerene and MWCNT under different conditions. Chemosphere. 139, 18-22.

Toropova A.P. Toropov A.A. Kudyshkin V.O. Rallo R. (2015): Prediction of the Q-e parameters from structures of transfer chain agents. J. Polym. Res. 22, 128.

Manganelli S., Leone C., Toropov A.A. Toropova A.P. Benfenati E. (2015): QSAR model for cytotoxicity of silica nanoparticles on human embryonic kidney cells. Materials Today: Proceedings (in press).
Sizochenko N., Rasulev B., Gajewicz A., Kuzmin V.E. Puzyn T., Leszczynski J. (2014): From basic physics to mechanisms of toxicity: Liquid Drop approach applied to develop predictive classification models for toxicity of metal oxide nanoparticles. Nanoscale. 6, 13986-13993.
Kar S., Gajewicz A., Puzyn T., Roy K. (2014): Periodic table-based descriptors to encode cytotoxicity profile of metal oxide nanoparticles: A mechanistic QSTR approach. Ecotoxicol. Environ. Safe. 107C, 162-169.
Toropov A.A. Toropova A.P. (2014): Optimal descriptor as a translator of eclectic data into endpoint prediction: Mutagenicity of fullerene as a mathematical function of conditions. Chemosphere. 104, 262-264.

Toropova A.P. Toropov A.A. Benfenati E., Puzyn T., Leszczynska D., Leszczynski J. (2014): Optimal descriptor as a translator of eclectic information into the prediction of membrane damage: The case of a group of ZnO and TiO2 nanoparticles. Ecotoxicol. Environ. Safe. 108, 203-209.

Worachartcheewan A., Mandi P., Prachayasittikul V., Toropova A.P. Toropov A.A. Nantasenamat C. (2014): Large-scale QSAR study of aromatase inhibitors using SMILES- based descriptors. Chemometr. Intell. Lab. Syst. 138, 120-126.
Gromelski M., Lewandowska W., Gajewicz A., Puzyn T. (2016): Comparative nano-QSAR modelling of cellular uptake for coated nanoparticles in pancreatic carcinoma cell line (PaCa2) based on calculations at the DFT and semi-empirical level of theory. Manuscript is being prepared - will be submitted until April 2016.
Lewandowska W., Gromelski M., Gajewicz A., Puzyn T. (2016): Towards understanding mechanisms of cellular uptake of superparamagnetic iron oxide nanoparticles (SPIONs) in human umbilical vein endothelial cells - local versus global nano-QSAR modelling. Manuscript is being prepared - will be submitted until April 2016.

Obtained results have demonstrated the potential benefits of using chemoinformatics approaches such as Nano-QSAR and Na no-QSPR modelling to obtain predictive knowledge for organic and inorganic nanoparticles that affect human cells and environment, and utilize this knowledge to improve the experimental design of NPs and enable their prioritization for in vivo testing. Presented methods allow for predictions of toxicity and relevant phys/chem endpoints of nanoparticles. Without doubts developed computational tools will open new opportunities to evaluate their activity as well as properties without necessity of performing expensive experimental studies on large pool of nanomaterials.

It's worth to highlight, that the very valuable output of Task 4.2 is to demonstrate the usefulness of methods of causal discovery to elucidate the underlying structure of the nanotoxicity data and retrieve additional, more robust interpretation for the developed SAR models. We have presented that the causal structures can efficiently be used in Nano-SAR modeling as additional criteria for quality evaluation. A great advantage of presented method is the fact that it provides the mechanistic interpretation of obtained results. In other words, methods of causal discovery open new opportunities to provide useful information at the molecular level that could be used to reveal mechanisms of toxicity. As case studies, the structure-activity relationships for toxicity of metal oxide nanoparticles (Nano-SARs) towards BEAS-2B and RAW 264.7 cell lines were established. To describe the nanoparticles, the simple ionic, fragmental and “liquid drop model” based descriptors that represent the nanoparticles' structure and characteristics were applied. The developed classification Nano-SAR models were validated to confirm reliability of predicting toxicity for all studied metal oxide nanoparticles. Developed models suggest different mechanisms of nanotoxicity for the two types of cells. For more details, please refer to:
Sizochenko N., Rasulev B., Gajewicz A., Mokshyna E., Kuzmin V.E. Leszczynski J., Puzyn T. (2015): Causal inference methods to assist in mechanistic interpretation of classification nano-SAR models. RSC Advances, 5, 77739-77745.

Another scientific paper, which demonstrates the usefulness of methods of causal discovery to elucidate the underlying structure of the nanotoxicity data is:

• Sizochenko N., Gajewicz A., Leszczynski J., Puzyn T. (2016): Causation or only correlation? Application of causal inference graphs for evaluating causality in nano-QSAR models. Nanoscale. Nanoscale. DOI: 10.1039/C5NR08279J.

Task 4.3: Comparing efficiency of CoMFA/CoMSIA and Hansch Analysis modelling schemes in Nano-QSAR

The efficiency and applicability of two the most widely applied QSAR approaches: Hansch analysis and CoMFA/CoMSIA have been compared. Classic Nano-QSAR and 3D Nano-QSAR were compared according to the following criteria: (i) efficiency, (ii) type of experimental data, (iii) type of nanomaterials, (iv) time required and computational cost and (v) software availability. The results of comparison can be summarised as follow:

1/ Experimental data:
Nano-QSAR - in vitro, in vivo
3D Nano-QSAR - in vitro: ligand-based response
2/ Nanomaterials:
homogeneity:
Nano-QSAR - inorganic, organic, metals - homogenous set
3D Nano-QSAR – organic - heterogeneous data with the same mode of action
data preparation:
Nano-QSAR - calculation of nanodescriptors
3D Nano-QSAR - establishment of the bioactive conformation of each molecule (docking)
3/ Statistics obtained:
Nano-QSAR - determination coefficients for calibration and validation, root means square errors
3D Nano-QSAR - determination coefficients for calibration and validation, root means square errors
4/ Time:
Nano-QSAR - limited by descriptor calculation
3D Nano-QSAR - limited by docking procedure
5/ Computational costs:
Nano-QSAR - limited by descriptor calculation
3D Nano-QSAR - limited by docking procedure
6/ Software:
Nano-QSAR - commercially available - user friendly
3D Nano-QSAR - commercially available - user friendly

Taking into account the advantages and limitations of both methods, we provided the recommendations for Nano-QSAR modellers to determine a proper methodology for investigation biological activity of nanoparticles. According to the mentioned recommendation on which approach (i.e. classic Nano-QSAR or 3D Nano-QSAR) should be applied in order to better understand the biological activity of nanomaterials requires the answer to the following questions:

What types of experimental data are considered?
What types of nanomaterials are considered?
What is the goal of the study?

It is obvious that adequate experimental data are essential to obtained proper models, both in case of classic Nano-QSAR and 3D approach. Appropriate data should fulfil two main principles: (i) should be measured according to the same protocol (ideally if they could be from the same source); and (ii) should be symmetrically distributed around their mean and their precision should be distributed over its range of variation. The more extensive discussion on biological data for nanomaterials one can find in the literature. Besides the quality of the data, the type of measured response is important to answer the first listed above question. In this point, it worth to emphasize that classic Nano-QSAR is more universal approach. There are models developed for particular molecular targets response, cell response, or the response measured on higher level of organism organization. On the other hand, in the 3D Nano-QSAR approach the receptor-based response is required, which in fact means that before 3D Nano-QSAR can be developed there should be the evidences of possible interactions between nanomaterials and the residues of the particular protein. This knowledge one can obtain directly by performing the proper experimental studies (e.g. X-ray crystallography, NMR) or indirectly by applying classic QSAR studies. Defining the type of descriptors that correlate with the modelled activity, in many cases allows finding the molecular target of this process. 3D Nano-QSAR model should not be developed unless one expects that the analysis will reveal insights into 3D interaction between ligand and receptor in its binding pocket.

Decision which approach can be applied depends also on the chemical nature of nanomaterials. There is no limitation in application of classic Nano-QSAR considering type of chemicals for which this model could be applied (organic, inorganic, metals, etc.), meanwhile 3D Nano-QSAR is rather applicable for organic nanomaterials
The third question is about the goal of the study. If the biological target is not known, and the objective is to find this target or gather general information about the biological activity of nanomaterials, the classic Nano-QSAR would be the right choice. But, if we know the binding pocket of the studied materials, the 3D Nano-QSAR might provide more useful information to understand this activity. This third question is purposely shown as the last issue in the experiment design flow, although the objectives should be rather defined on the first stage of the project. However, QSAR modellers are in most cases rather the users of already acquired experimental data then the researchers who are employed in the experiment design. Thus, they are forced to define goals considering the availability of data. To overcome these difficulties, stronger collaboration between experimentalists and modellers is recommended.

A detailed description of this work has been presented in deliverable D4.3 as well as in: Jagiello K., Grzonkowska M., Swirog M., Ahmed L., Rasulev B., Avramopoulos A., Papadopoulos M., Leszczynski J., Puzyn T. (2016): On the applicability of 3D QSAR in nanoQSAR studies. Manuscript is being prepared and will be submitted until April 2016. In the mentioned above contribution, we stated that both approaches, Nano-QSAR and 3D Nano-QSAR, could be used simultaneously, if it is possible. Application of classic Nano-QSAR model, which is more universal approach, would allow gathering general information about the mode of biological activity of nanomaterials. Then, the 3D QSAR application would help in understanding this activity in detail.

Task 4.4: Estimating the environmental behaviour of NPs based on the phys/chem data predicted with Nano-QSAR

Rapid developments of engineered nanomaterials raise concern regarding their potential adverse on the environment and human health. Predicting the environmental impact of engineered nanomaterials requires estimating potential level of environmental exposure concentration of engineered nanomaterials in the various media such as air, water, soil, and vegetation. However, due the lack of data and knowledge regarding: (i) emission rates; (ii) physicochemical properties (i.e. size distribution, surface reactivity, state of purity), as well as (iii) processes occurring with environmental compartments (i.e. aggregation, dissolution, and attachment mechanisms) estimation the environmental behaviour of NPs remains speculative. Thus, during reporting period, relatively simple environmental fate model that uses first-order kinetics to estimate environmental background concentrations for engineered nanomaterials in an environmental system being composed of the compartments (i.e. air, soil, water, and sediment) which are represented as boxes linked together via mechanistic intermedia transport processes were discussed. We have critically reviewed the possibility of modelling environmental transport and fate of engineered nanoparticles. In addition we have defined the specific properties and processes that affect the fate and behaviour of NMs in environment, focusing on the specific properties of NMs that modulate the interactions in the environment.

A detailed description of this work has been presented in deliverable D4.4 as well as in manuscript that has been submitted: Sikorska C. Puzyn T. (2016): Multimedia modelling of engineered nanoparticles: a review of fate and transport studies. Submitted.

Task 4.5: Investigating the minimum requirements sufficient for successful validation of models according to OECD

We proposed - in collaboration with other EU FP7 projects focused on developing tools for computational risk assessment of nanoparticles (MODERN, Mod-Enp-Tox, PreNanoTox, MembraneNanoPart, eNanoMapper) - an interpretation of well-known “OECD Principles for (Q)SAR validation” in the context of nano-(Q)SAR and presented our opinion on the criteria to be fulfilled by every model developed for nanoparticles.

A detailed description of this work has been presented in deliverable D4.5 as well as in manuscript that has been submitted: Puzyn T., Jeliazkova N, Sarimveis H., Marchese Robinson R. L., Lobaskin V., Rallo R., Richarz A. N., Gajewicz A., Papadopulos M. G., Hastings J., Cronin M. T. D., Benfenati E., Fernandez A. (2016): On the validation criteria of (Q)SAR models used in nanotechnology. Submitted. It should be highlighted that the preparation of a joint paper has been initiated and coordinated by NanoPUZZLES project.

In the mentioned above contribution, we stated that (Quantitative) Structure-Activity Relationship ([Q]SAR) modelling creates an interesting option among the available methods of risk assessment. And it might be used for predicting a variety of properties, including toxicity of newly designed nanoparticles. However, since every (Q)SAR model must be appropriately validated, which is crucial for ensuring accuracy of the predictions we presented our opinion on the criteria to be fulfilled by every model developed for nanoparticles. In general, we agreed that the OECD Principles for (Q)SAR validation creates an appropriate framework for validating nano-(Q)SARs as well. However, special attention should be given for some issues specific for nanoparticles. The most important conclusions are as follows:
Since the ontology for studying safety of nanoparticles is still under development, the modelled endpoint requires a careful consideration and detailed description, while reporting the model.
Classic molecular descriptors might be inappropriate for modeling nanoparticles. Thus, they should be used with criticism. On the other hand, newly developed descriptors should be validated and reported with providing details necessary to calculate them by anyone interested.
It is highly recommended presenting the models in Predictive Model Markup Language (PMML) format and, whenever possible, placing them in public repositories to ensure their transparency and reproducibility.
Compared to classical (Q)SAR modelling, it seems much more important to achieve a balance between the level of confidence in the predictions of a nano-(Q)SAR model and the scope (applicability domain) of the model (balance between “local” and “global” models).
Nano-bio interactions involving nanoparticles are not fully determined by the particle chemistry alone. Because of that, mechanistic interpretation of nano-(Q)SARs can be more problematic than that for classic (Q)SAR models; very often a wider context is required.

Task 4.6: Development of the conceptual framework for further grouping of NPs

Recently, there has been an ongoing discussion on how to develop scientifically based categorization strategies? how to classify and prioritise NMs for safety assessment? or how to define the key physico-chemical features and toxicological responses allowing the effect-driven grouping of nanomaterials? The overview on the existing concepts, schemes as well as criteria for the grouping of nanomaterials, including classification based on:
(1) commercial/economic importance;
(2) production volume and value;
(3) chemical composition;
(4) intrinsic properties, such as: dimensionality, shape, morphology, surface, aspect ratio;
(5) complexity and functionality of nanoparticles;
(6) extrinsic properties, such as: biopersistence, biokinetics, environmental parameters
(7) modes of action;
(8) parameters with a proven link to toxicological or biological response(s), i.e. the adverse outcome pathways.14 However, on the other hand, our knowledge regarding the harmful interactions of engineered nanoparticles with biological systems, as well as with the environment, is still insufficient and very limited. It is still largely unknown which properties determine and/or influence the toxicity of particular nanoparticles. Also, it is not possible to draw any general conclusions or develop mechanisms of the potential toxicity of nanomaterials.

Therefore, taking into account the current state of the art in nanoscience and nanotoxicology development, within the NanoPUZZLES project we have proposed the following framework for grouping NPs:
grouping/ranking nanomaterials using chemometric techniques such as: principal component analysis (PCA), hierarchical cluster analysis (HCA) based on: (i) variation in chemical structure and physico-chemical properties (represented by experimental and theoretical descriptors) as well as (ii) possible mechanisms of metabolism and/or mode of action (represented by experimental endpoints(s));
applying similarity analysis using for example hierarchical cluster analysis (HCA) to: (i) verifying of the homogeneity of the group, which was defined based on grouping approaches/criteria as well as (ii) verifying, whether the data that are available for the selected group are sufficiently representative, whether the group should be divided into smaller subgroups/subcategories;
identifying of data gaps in physicochemical characterization, exposure assessment, hazard assessment within the defined group(s);
applying modelling approaches (i.e. computational methods such as QSAR/QSPR and read-across) for predicting missing data for specific NPs within the defined group(s);
using research outcomes for the prioritization of hazardous, if possible.

A detailed description of this work has been presented in deliverable D4.6. In addition, following manuscript which summarizes all grouping efforts and correlates the results of grouping by structure with grouping by properties is currently being prepared and will be submitted by the end of March 2016: Gajewicz A., Marchese Robinson R.L. Cronin M.T.D. Puzyn T. (2016): How to overcome problem of insufficient empirical data for nanomaterials? Grouping and read-across studies.
It's worth to highlight, that the very valuable output of Task 4.6 is novel and effective algorithm for filling missing data in quantitative manner when the number of experimental data is insufficient to develop appropriate Nano-QSAR model. Within the NanoPUZZLES project the methodology of quantitative read-across for nanomaterials has been developed. The main idea standing behind on the read-across algorithm refers to the assumption that two NPs are similar when located close to each other (in terms of a measure of distance). Euclidean distance was proposed to be used as similarity measure (to search the similarities (and dissimilarities) of the studied compounds) and then data gaps can be estimated by simple scaling the available experimental data from source chemicals (i.e. two compounds that were located closest to the target compound in terms of a measure of distance) to the target compound by using the two-point slope formula. The equation that goes through the two given points (NP_1, NP_2) represented as (x1, y1) and (x2, y2) respectively, can be computed according to following formula:


(y0-y2)/(x0-x2)=(y1-y0)/(x1-x0) -> y0=[y1(x0-x2)+y2(x1-x0)]/(x1-x2)


where: xi is a particular value of a given independent variable for NP_i; yi is the value of the endpoint for NP_i.

Proposed approach provides a capable tool allowing for predictions of various properties of unknown nanomaterials based on information extracted from very few known species with a similar level of accuracy as provided by Nano-QSAR model(s). It opens new opportunities to evaluate their properties without necessity of performing expensive experimental studies on large pool of nanomaterials.

Additionally to facilitate grouping and read across modeling the following software tools have been developed: (1) Clustering - Modified k-Medoid (http://nanobridges.eu/wp-content/uploads/2014/11/modifiedKMedoid-Manual.pdf) and (2) NanoProfiler (http://nanobridges.eu/wp-content/uploads/2014/11/NanoProfiler-Manual.pdf). It needs to be highlighted, however that the development freely available on-line/standalone nano-profiler was performed under the NanoBRIDGES project, in collaboration with NanoPUZZLES. We specify that the work undertaken by the Indian authors have been funded within NanoBRIDGES, while the contribution from the project partners was funded by NanoPUZZLES. The contribution done within NanoPUZZLES refers to the development of algorithm(s), which can be applied to perform grouping as well as read-across of nanoparticles, whereas the translation the algorithm into programming language and preparation the final software has been done within NanoBRIDGES project. Software’s developed within NanoBRIDGES project along with their manuals are freely available on the project webpage: http://nanobridges.eu/software/. In addition for a better dissemination of the developed software tools, following paper has been published: Ambure P., Aher R.B. Gajewicz A., Puzyn T., Roy K. (2015): "NanoBRIDGES" software: Open access tools to perform QSAR and nano-QSAR modeling. Chemometr. Intell. Lab. Syst. 147, 1-13.

References:

1) http://nbi.oregonstate.edu/

2) Pathakoti K., Huang M.-J. Watts J.D. He X., Hwang H.-M. (2014): Using experimental data of Escherichia coli to develop a QSAR model for predicting the photo-induced cytotoxicity of metal oxide nanoparticles. Journal of Photochemistry and Photobiology B: Biology, 130, 234-240.

3) Golbamaki N., Rasulev B., Cassano A., Marchese Robinson R.L. Benfenati E., Leszczynski J., Cronin M.T.D. (2015): Genotoxicity of metal oxide nanomaterials: review of recent data and discussion of possible mechanisms. Nanoscale, 7, 2154-2198.

4) Marchese Robinson R.L. Lynch I. Peijnenburg W., Rumble J., Klaessig F., Hendren C.O. Marquardt C., Rauscher H., Puzyn T., Purian R., Åberg C., Karcher S., Vriens H., Hoet P., Hoover M.D. Harper S.L. (2016): How should the completeness and quality of curated nanomaterial data be evaluated? Nanoscale (article under review).

5) http://biocenitc-deq.urv.cat/nanodms

6) http://www.myexperiment.org/files/1356.html

7) https://github.com/RichardLMR/xls2txtISA.NANO.archive

8) Marchese Robinson R.L. Richarz A.N. Cassano A., Cronin M.T.D. Rallo R. (2015): An ISA-TAB-Nano based data collection framework to support data-driven modelling of nanotoxicology. Beilstein Journal of Nanotechnology, 6, 1978-1999.

9) http://www.nanopuzzles.eu/results-products

10) http://dx.doi.org/10.5281/zenodo.35493

11) http://dx.doi.org/10.5281/zenodo.35419

12) https://figshare.com/search?q=NanoPUZZLES+project&quick=1 N.B. Links to the FigShare versions of these datasets are provided via the cited Zenodo records containing copies of all datasets [http://dx.doi.org/10.5281/zenodo.35493

13) http://dx.doi.org/10.5281/zenodo.35419].

14) Lynch I., Weiss C., Valsami-Jones E. (2014): A strategy for grouping of nanomaterials based on key physico-chemical descriptors as a basis for safer-by-design NMs. Nano Today, 9, 3:266–270.

Potential Impact:
NPs are often very complex chemical systems, therefore only comprehensive and interdisciplinary analyses may deliver new reliable knowledge on their properties and (eco)toxicity. NanoPUZZLES project integrates knowledge and methods between various disciplines: materials science, toxicology and ecotoxicology, quantum chemistry and molecular mechanics as well as chemoinformatics. This results in increasing comprehensiveness of studying the risk related to introducing nanomaterials.

The main objective of the NanoPUZZLES project is to deliver a package of algorithms for the modelling of relationships between the structure, properties, molecular interactions and toxicity of engineered nanoparticles (NPs) that can be applied within industry to design safe and environmentally friendly nanomaterials. This “package of algorithms” includes the following:
• quality assesses data sets of experimental values (physico-chemical properties and toxicity) for NPs, for modelling,
• tools for calculating descriptors of nanoparticles' structures,
• computational protocols for studying the interactions of NPs with biological systems,
• new grouping schemes for nanoparticles, based on their structure, physico-chemical properties, and toxicity,
• a basis for quantitative modelling of the relationships between the structure and physic-chemical properties and toxicity of NPs as well as preliminary nano-QSAR models for selected endpoints.

The outputs of the NanoPUZZLES project - algorithms, reports, databases, software tools, publications, articles and conference presentations - have a significant impact on various groups in society, including scientists, administration, industry and citizens. However, actions undertaken in each work package have different influences on each of the respective target groups. The first two work packages (NanoDATA and NanoDESC) have direct effects on scientists as they enhance the general understanding of nanoparticles structure and properties, whereas interactions of NPs with human health and the environment (Work Package 3: NanoINTER) is the most important for end users of products containing NPs – the public. Industry (mostly pharmaceutical and biomedical, but also food, telecommunications and electronics) will benefit the most from practical outputs of the project, which are contained in Work Package 4: NanoQSAR.

Without doubts, NanoPUZZLES project directly influences the scientific community by providing new models and methods for testing of the safety of nanomaterials. Application of these methods and consideration of the developed data collection provide information on the intrinsic properties of the NPs that lead to an understanding of their behaviour in physical processes in the environment and to the development of relationships between physicochemical properties and biological activity/toxicity.

It does also affect the authorities by providing knowledge-driven recommendations. Moreover, employing computational techniques developed within the NanoPUZZLES project should significantly reduce the time and cost of risk assessment for novel engineered nanoparticles. The reduction of costs related to risk assessment would increase the probability of success of the REACH system and other regulations (e.g. the Cosmetic Directive) that implement the European policy of the 3Rs (Replacement, Reduction and Refinement of animal use).

NanoPUZZLES outputs also directly influence industry by providing new technologies for the production of safe nanomaterials. Since computational techniques developed within the NanoPUZZLES allow to design nanomaterials in an environmentally friendly way that will be of low risk for human health and the environment while being competitive on the market (according to the idea of “green marketing”) thus the outputs of the NanoPUZZLES project will ultimately have numerous applications in industry and environmental hazard assessment - guiding reliable risk and safety evaluations for these materials to ensure their safety for human health and the environment.

It is expected that the largest impact will be observed for citizens as potential consumers of engineered nanomaterials. They will be provided with new, safe products and informed by authorities about their safe use, application and contraindications.

During the project implementation stage, the particular target groups (i.e. scientists, authorities, industry and citizens) were reached via appropriate measures presented below:

1/ Scientists (S):
• Targeted stakeholders within the group - Scientists developing methods of risk assessment (empirical and/or computational);
• Communication channel - On-line materials, conferences, training sessions, publications;
• Aim to be achieved within this group – (1) Promoting the idea of use computational techniques instead of empirical testing; (2) Teaching new methods dedicated to risk assessment of engineered nanoparticles; (3) Inspiring a wider group of scientists to work on development of those methods; (4) Receiving scientific feedback on new knowledge and methods developed within the project; (5) Comparing the obtained results with the results of competing research projects.

2/ Administration/Health and Environmental Authorities (A):
• Targeted stakeholders within the group - persons and organisations responsible for planning regional strategies of economic and scientific/ technological development and overseeing regulatory acceptance of nanomaterial containing products;
• Communication channel - On-line materials, conferences;
• Aim to be achieved within this group – (1) To set new laws and regulations for the production of new nanomaterial containing products; (2) To encourage regulatory acceptance of computational methods in risk assessment of nanomaterials.

3/ Business/Industry designers (I):
• Targeted stakeholders within the group - Business/Industry designers and producers of novel nanomaterials, usually not familiar with the necessity and/or the procedures
of risk assessment;
• Communication channel - On-line materials, conferences, dissemination (fairs, e-mails, newsletters);
• Aim to be achieved within this group – (1) To manufacture safe nanomaterial containing products; (2) To reduce the costs associated with toxicity testing and development of new nanomaterial containing products.

4/ Citizens (C):
• Targeted stakeholders within the group - citizens, who will benefit by using new, safe and verified products with known toxicity and predicted physico-chemical properties;
• Communication channel - On-line materials;
• Aim to be achieved within this group – (1) Communication with stakeholders outside the network and the public at large, for the purpose of spreading the information and dissemination of knowledge will be carried out extensively during the whole project duration.

The major tools for exploiting the results of the NanoPUZZLES project are presented below:

1) NanoPUZZLES project website - the website plays the role of an arena for exchange of information, where information about the work done within the project will be exhibited.

2) Publications in scientific journals - the following publications were acknowledged to the NanoPUZZLES project during the reporting period:

Scientific papers – published:
• Sizochenko N., Gajewicz A., Leszczynski J., Puzyn T. (2016): Causation or only correlation? Application of causal inference graphs for evaluating causality in nano-QSAR models. Nanoscale. Nanoscale. DOI: 10.1039/C5NR08279J.
• Avramopoulos A., Otero N., Karamanis P., Pouchan C., Papadopoulos M. G. (2016): A Computational Study of the Interaction and Polarization Effects of Complexes involving Molecular Graphene and C60 or a Nucleobasis. J. Phys. Chem. A. DOI: 10.1021/acs.jpca.5b09813.
• Manganelli S., Leone C., Toropov A.A. Toropova A.P. Benfenati E. (2016): QSAR model for predicting cell viability of human embryonic kidney cells exposed to SiO2 nanoparticles. Chemosphere. 144, 995-1001.

• Marchese Robinson R.L. Richarz A.N. Cassano A., Cronin M.T.D. Rallo R. (2015): An ISA-TAB-Nano based data collection framework to support data-driven modelling of nanotoxicology. Beilstein J, Nanotechnol. 6, 1978-1999.
• Golbamaki N., Rasulev B., Cassano A., Marchese Robinson R.L. Benfenati E., Leszczynski J., Cronin M.T.D. (2015): Genotoxicity of metal oxide nanomaterials: review of recent data and discussion of possible mechanisms. Nanoscale. 7, 2154-2198.
• Sizochenko N., Rasulev B., Gajewicz A., Mokshyna E., Kuzmin V.E. Leszczynski J., Puzyn T. (2015): Causal inference methods to assist in mechanistic interpretation of classification nano-SAR models. RSC Advances. 5, 77739-77745.
• Sizochenko N., Jagiello K., Leszczynski J., Puzyn, T. (2015): How the “Liquid Drop” approach could be efficiently applied for Quantitative Structure–Property Relationship modeling of nanofluids. J. Phys. Chem. C, 119, 25542–25547.
• Ambure P., Aher R.B. Gajewicz A., Puzyn T., Roy K. (2015): "NanoBRIDGES" software: Open access tools to perform QSAR and nano-QSAR modeling. Chemometr. Intell. Lab. Syst. 147, 1-13.
• Jagiello K., Puzyn T. (2015): Computational techniques application in environmental exposure assessment. In: Quantitative structure-activity relationships in drug design, predictive toxicology, and risk assessment. Ed. Kunal Roy, IGI Global, USA, pp. 471-505 ISSN: 2327-5448.
• Leonis G., Avramopoulos A., Papavasileiou K. D., Reis H., Steinbrecher T., Papadopoulos M. G. (2015): A Comprehensive Computational Study of the Interaction between Human Serum Albumin and Fullerenes. J. Phys. Chem. B, 119, 14971-14985.
• Vrontaki E., Leonis G., Avramopoulos A., Papadopoulos M. G., Simcic M., Grdadolnik S. G., Afantitis A., Melagraki G., Hadjikakou S. K., Mavromoustakos T. (2015): Stability and binding effects of silver(I) complexes at lipoxygenase-1. J Enzyme Inhib. Med. Chem. 30, 539-549.
• Tzoupis H., Leonis G., Avramopoulos A., Reis H., Czyżnikowska Ż., Zerva S., Vergadou N., Peristeras L. D., Papavasileiou K. D., Alexis M. N., Mavromoustakos T., Papadopoulos M. G. (2015): Elucidation of the binding mechanism of renin using a wide array of computational techniques and biological assays. J. Mol. Graph. Model. 62, 138-149.
• Duchowicz P.R. Fioressi S.E. Bacelo D.E. Saavedra L.M. Toropova A.P. Toropov A.A. (2015): QSPR Studies on Refractive Indices of Structurally Heterogeneous Polymers. Chemometr. Intell.Lab. Syst. 140, 86–91.

• Richarz A.N. Avramopoulos A., Benfenati E., Gajewicz A., Leonis G., Marchese Robinson R.L. Papadopoulos M.G. Cronin M.T.D. Puzyn T. (2015): The EU NanoPUZZLES Project. Modelling the Toxicity of Nanoparticles, Springer (Book chapter in press).
• Veselinovic J. B., Toropov A. A., Toropova A.P. Nikolic G.M. Veselinovic A.M. (2015): Monte Carlo Method-Based QSAR Modeling of Penicillins Binding to Human Serum Proteins. Arch. Pharm. Chem. Life Sci. 348, 62-67.

• Toropova A.P. Toropov A.A. Rallo R., Leszczynska D., Leszczynski J. (2015): Optimal descriptor as a translator of eclectic data into prediction of cytotoxicity for metal oxide nanoparticles under different conditions. Ecotoxicol. Environ. Safe. 112, 39-45.

• Toropov A.A. Toropova A.P. (2015): Quasi-QSAR for mutagenic potential of multi-walled carbon-nanotubes. Chemosphere. 124, 40–46.

• Toropova A.P. Toropov A.A. (2015): Mutagenicity: QSAR - quasi-QSAR - nano-QSAR. Mini-Reviews in Medicinal Chemistry, 15, 608-621.

• Toropova A.P. Toropov A.A. Veselinovic J. B., Veselinovic A.M. Benfenati E., Leszczynska D., Leszczynski J. (2015): Application of the Monte Carlo method to prediction of dispersibility of graphene in various solvents. Int. J. Environ. Res. 9, 1211-1216.

• Toropov A.A. Toropova A.P. Veselinovic A.M. Veselinovic J. B., Nesmerák K., Raska Jr I., Duchowicz P. R., Castro E. A., Kudyshkin V.O. Leszczynska D., Leszczynski J. (2015): The Monte Carlo method based on eclectic data as an efficient tool for predictions of endpoints for nanomaterials - two examples of application. Com. Chem. High T. Scr. 18, 376-386.

• Toropova A.P. Toropov A.A. (2015): Quasi-SMILES and nano-QFAR: United model for mutagenicity of fullerene and MWCNT under different conditions. Chemosphere. 139, 18-22.

• Toropova A.P. Toropov A.A. Kudyshkin V.O. Rallo R. (2015): Prediction of the Q-e parameters from structures of transfer chain agents. J. Polym. Res. 22, 128.

• Manganelli S., Leone C., Toropov A.A. Toropova A.P. Benfenati E. (2015): QSAR model for cytotoxicity of silica nanoparticles on human embryonic kidney cells. Materials Today: Proceedings (in press).

• Richarz A.N. Madden J.C. Robinson L.M.R. Lubiński L., Mokshina E., Urbaszek P., Kuz'min V.E. Puzyn T., Cronin M.T.D. (2015): Development of computational models for the prediction of the toxicity of nanomaterials. Perspectives in Science. 3, 27-29.
• Sizochenko N., Rasulev B., Gajewicz A., Kuzmin V.E. Puzyn T., Leszczynski J. (2014): From basic physics to mechanisms of toxicity: Liquid Drop approach applied to develop predictive classification models for toxicity of metal oxide nanoparticles. Nanoscale. 6, 13986-13993.
• Leonis G., Avramopoulos A., Salmas R. E., Durdagi S., Yurtsever M., Papadopoulos M. G. (2014): Elucidation of Conformational States, Dynamics, and Mechanism of Binding in Human κ-Opioid Receptor Complexes. J. Chem. Inf. Model. 54, 2294–2308.
• Tzoupis H., Leonis G., Avramopoulos A., Mavromoustakos T., Papadopoulos M. G. (2014): Systematic molecular dynamics, MM-PBSA, and ab initio approaches to the saquinavir resistance mechanism in HIV-1 PR Due to 11 double and multiple mutations. J. Phys. Chem. B, 118, 9538-9552.
• Kar S., Gajewicz A., Puzyn T., Roy K. (2014): Periodic table-based descriptors to encode cytotoxicity profile of metal oxide nanoparticles: A mechanistic QSTR approach. Ecotoxicol. Environ. Safe. 107C, 162-169.
• Toropov A.A. Toropova A.P. (2014): Optimal descriptor as a translator of eclectic data into endpoint prediction: Mutagenicity of fullerene as a mathematical function of conditions. Chemosphere. 104, 262-264.

• Toropova A.P. Toropov A.A. Benfenati E., Puzyn T., Leszczynska D., Leszczynski J. (2014): Optimal descriptor as a translator of eclectic information into the prediction of membrane damage: The case of a group of ZnO and TiO2 nanoparticles. Ecotoxicol. Environ. Safe. 108, 203-209.

• Worachartcheewan A., Mandi P., Prachayasittikul V., Toropova A.P. Toropov A.A. Nantasenamat C. (2014): Large-scale QSAR study of aromatase inhibitors using SMILES- based descriptors. Chemometr. Intell. Lab. Syst. 138, 120-126.

• Richarz A.N. Cronin M.T.D. Madden J., Lubinski L., Mokshina E., Urbaszek P., Puzyn T., Kuz’min V. (2013): Toxicity of nanomaterials: Availability and suitability of data for the development of in silico models. Toxicology Letters 221, 246.
• Leonis G., Steinbrecher T., Papadopoulos M. G. (2013): A Contribution to the Drug Resistance Mechanism of Darunavir, Amprenavir, Indinavir, and Saquinavir Complexes with HIV-1 Protease Due to Flap Mutation I50V: A Systematic MM–PBSA and Thermodynamic Integration Study. J. Chem. Inf. Model. 53, 2141-2153.

Scientific papers – submitted (under review):
• Puzyn T., Jeliazkova N, Sarimveis H., Marchese Robinson R. L., Lobaskin V., Rallo R., Richarz A. N., Gajewicz A., Papadopulos M. G., Hastings J., Cronin M. T. D., Benfenati E., Fernandez A. (2016): On the validation criteria of (Q)SAR models used in nanotechnology. Submitted.
• Marchese Robinson R.L. Lynch I., Peijnenburg W., Fritts M., Rumble J., Klaessig F., Hendren C.O. Marquardt C., Rauscher H., Puzyn T., Purian R., Åberg C., Karcher S., Gaheen S., Vriens H., Hoet P., Hoover M.D. Ku S., Harper S. (2016): How should the completeness and quality of curated nanomaterial data be evaluated? Submitted.
• Odziomek K., Ushizima D., Oberbek P., Kurzydłowski K. J., Puzyn T., Haranczyk M. (2016): Scanning electron microscopy image representativeness: morphological data on nanoparticles. Submitted
• Papavasileiou K., Avramopoulos A., Leonis G., Papadopoulos M. (2016): Computational Investigation of Fullerene DNA Interactions: Implications of Fullerene’s Size, and Functionalization on DNA Structure and Binding Energetics. Submitted.

Scientific papers – will be submitted until April 2016:
• Gajewicz A. (2016): How to overcome problem of insufficient empirical data for nanomaterials? Read-across studies. Manuscript is being prepared.
• Jagiello K., Grzonkowska M., Swirog M., Ahmed L., Rasulev B., Avramopoulos A., Papadopoulos M., Leszczynski J., Puzyn T. (2016): On the applicability of 3D QSAR in nanoQSAR studies. Manuscript is being prepared.
• Jagiello K., Chomicz B., Avramopoulos A., Gajewicz A., Mikolajczyk A., Bonifassi P., Papadopoulos M., Leszczynski J., Puzyn T. (2016): Size-dependent electronic properties of nanomaterials – How this new class of nanodescriptors supposed to be calculated? Manuscript is being prepared.
• Sizochenko N., Syzochenko M., Gajewicz A., Leszczynski J., Puzyn T. (2016): Application of Feature Net Approach to Model Thermal Conductivity and Viscosity of Nanofluids. Manuscript is being prepared.
• Golbamaki N., Golbamaki A., Sizochenko N., Rasulev B., Cassano A., Marchese Robinson R., Cronin M.T.D. Puzyn T., Leszczynski J., Benfenati E. (2016): Classification Nano-SAR modeling of metal oxides nanoparticles genotoxicity based on Comet assay data. Manuscript is being prepared.
• Gromelski M., Lewandowska W., Gajewicz A., Puzyn T. (2016): Comparative nano-QSAR modelling of cellular uptake for coated nanoparticles in pancreatic carcinoma cell line (PaCa2) based on calculations at the DFT and semi-empirical level of theory. Manuscript is being prepared.
• Lewandowska W., Gromelski M., Gajewicz A., Puzyn T. (2016): Towards understanding mechanisms of cellular uptake of superparamagnetic iron oxide nanoparticles (SPIONs) in human umbilical vein endothelial cells - local versus global nano-QSAR modelling. Manuscript is being prepared.

3) Conferences including NanoPUZZLES project conferences, international conferences, workshops, seminars, trade fairs - NanoPUZZLES project was presented during various external conferences listed below:

Conference presentations:
• Gajewicz A., Puzyn T. (2015): Insufficient empirical data do not have to create barriers for regulating nanomaterials: Applying in silico methods 23nd Conference on Current Trends in Computational Chemistry (23nd CCTCC), Jackson, USA, 13-14 November 2015 (oral presentation).
• Sizochenko N., Rasulev B., Gajewicz A., Leszczynski J., Puzyn T. (2015): Adding new features for nano-QSAR modeling: causal inference methods and mechanistic interpretation. 23rd Conference on Current Trends in Computational Chemistry (23rd CCTCC), Jackson, USA, 12-13 November 2015 (poster presentation).
• Mikolajczyk A., Puzyn T. (2015): Nano-QSAR approach towards development of efficient and safe photocatalysts based on Me@TiO2 nanoparticles. 23nd Conference on Current Trends in Computational Chemistry (23nd CCTCC), Jackson, USA, 13-14 November 2015 (poster presentation).
• Mikolajczyk, A. Rasulev, B., Pinto, H., Gajewicz, A., Leszczynski, J., Puzyn T. (2015): Computational modelling - how to design environmentally friendly photocatalyst based on Me@TiO2? 23rd Conference on Current Trends in Computational Chemistry (23nd CCTCC), Jackson, USA, 13-14 November 2015 (poster presentation).
• Leonis G. (2015): Modeling the Interactions Between Nanoparticles and Biomolecules. European Conference on Computational Nanotoxicology (CompNanoTox2015), Malaga, Spain, 4-6 November 2015 (oral presentation).
• Avramopoulos A. (2015): Quantum mechanical simulations for the study of the interactions between Nanoparticles and Biological Systems. European Conference on Computational Nanotoxicology (CompNanoTox2015), Malaga, Spain, 4-6 November 2015 (oral presentation).
• Gajewicz A., Puzyn T. (2015): Computational modelling - How to overcome problem of insufficient empirical data for nanomaterials? European Conference on Computational Nanotoxicology (CompNanoTox2015), Malaga, Spain, 4-6 November 2015 (oral presentation).
• Marchese Robinson R.L. Richarz A.N. Cassano A., Cronin M.T.D. Rallo R. (2015): Data collection from the nanotoxicology literature using ISA-TAB-Nano. European Conference on Computational Nanotoxicology (CompNanoTox2015), Malaga, Spain, 4-6 November 2015 (oral presentation).
• Cassano A., Marchese Robinson R.L. Richarz A.N. Cronin M.T.D. (2015): Modelling the in vitro cytotoxicity of metal/metal oxide and silica nanomaterials under diverse experimental conditions. European Conference on Computational Nanotoxicology (CompNanoTox2015), Malaga, Spain, 4-6 November 2015 (poster presentation).
• Toropova A.P. Toropova A.A. Benfenati E. (2015): Modelling nanomaterials with CORAL. CompNanoTox2015. Malaga, Spain, 4-6 November 2015 (poster presentation).
• Jagiello K., Grzonkowska M., Swirog M., Ahmed L., Rasulev B., Avramopoulos A., Leszczynski J., Puzyn T. (2015): The applicability of QSAR and 3D-QSAR approaches in Nano-QSAR studies. European Conference on Computational Nanotoxicology (CompNanoTox2015), Malaga, Spain, 4-6 November 2015 (poster presentation).
• Mikolajczyk, A., Malankowska, A., Gajewicz, A., Hirano, S., Zaleska-Medynska, A., Puzyn T. (2015): Design of new efficient and environmentally friendly photocatalysts: Application of nano-QSPR and mixture descriptors. European Conference on Computational Nanotoxicology (CompNanoTox2015), Malaga, Spain, 4-6 November 2015 (poster presentation).
• Barycki M., Puzyn T. (2015): Computational modelling of aggregation and sedimentation of nanoparticles based on the population balance equation. European Conference on Computational Nanotoxicology (CompNanoTox2015), Malaga, Spain, 4-6 November 2015 (poster presentation).
• Jagiello K., Chomicz B., Avramopoulos A., Papadopoulos M., Puzyn T. (2015): Size-dependent properties of nanomaterials – the new class of nanodescriptors. European Conference on Computational Nanotoxicology (CompNanoTox2015), Malaga, Spain, 4-6 November 2015 (poster presentation).
• Gromelski M., Lewandowska W., Gajewicz A., Puzyn T. (2015): Is more expensive always better? Comparison of different computational methods in nano-QSAR modelling of cellular uptake for nanoparticles. European Conference on Computational Nanotoxicology (CompNanoTox2015), Malaga, Spain, 4-6 November 2015 (poster presentation).
• Lewandowska W., Gromelski M., Gajewicz A., Puzyn T. (2015): Nano-QSAR modeling for cellular uptake of superparamagnetic iron oxide nanoparticles (SPIONs) in human umbilical vein endothelial cells (HUVEC). European Conference on Computational Nanotoxicology (CompNanoTox2015), Malaga, Spain, 4-6 November 2015 (poster presentation).
• Wyrzykowska E., Puzyn T., (2015): The assessment of existing methods in determining applicability domain of the model in the nano-QSPR approach. European Conference on Computational Nanotoxicology (CompNanoTox2015), Malaga, Spain, 4-6 November 2015 (poster presentation).
• Puzyn T. (2015): Computational Modeling in Nanotoxicology. 7th International Conference (NANOCON 2015), Brno, Czech Republic, 14-16 October 2015 (oral presentation).
• Jagiello K., Zaborowska M., Ahmed L., Avramopoulos A., Puzyn T. (2015): The applicability of CoMFA/CoMSIa approach in Nano-QSAR studies. 58 Zjazd Naukowy Polskiego Towarzystwa Chemicznego, Gdansk, Polska, 21-25 September 2015 (poster presentation).
• Mikolajczyk, A., Malankowska, A., Gajewicz, A., Zaleska-Medynska, A., Puzyn T. (2015): Zastosowanie metod komputerowych w projektowaniu bezpiecznych i funkcjonalnych układów fotokatalitycznych typu Men@MeOx w oparciu o zasadę „safe-by-design. 58th Annual Scientific Meeting of the Polish Chemical Society (PTChem), Gdansk, Poland, 21-25 September 2015 (poster presentation).
• Mikolajczyk, A., Pinto, H., Gajewicz, A., Puzyn, T., Leszczyński J. (2015): Nanoparticle characterizoation: NanoQSPR modeling of zeta potential (ζ) for metal oxide nanoparticles. 11th International Symposium on Recent Advances in Environmental Health Research, 13th International Symposium on Metal Ions in Biology & Medicine, Jackson, MS, USA, 13-16 September 2015 (poster presentation).
• Puzyn T. (2015): Achievements and perspectives of computational nanotoxicology. 51 Congress of the European Societies of Toxicology (EUROTOX2015), Porto, Portugal, 13-16 September 2015 (oral presentation).
• Avramopoulos A. (2015): Simulating interactions between nanoparticles and biological systems with the aid of quantum chemistry. 8th Swedish-Hellenic Life Science Research Conference, Athens, Greece, 12-13 September 2015 (oral presentation).
• Sizochenko N., Rasulev B., Puzyn T., Leszczynski J. (2015): The “4 C” of genotoxicity: Computation, Classification, Clustering and Causality. 15th Southern School on Computational Chemistry & Materials Science (SSCCMS), Jackson, USA, 23-24 July 2015 (poster presentation).
• Manganelli S. (2015): L'importanza delle tematiche ambiente e salute nella programmazione europea: l'esperienza dell'Istituto Mario Negri. Workshop nazionale: Scienza, tecnologia, salute umana ed ambiente: prospettive e opportunità nell’ambito della programmazione Horizon 2020. Lecce, Italy, 15 July 2015 (oral presentation).
• Gajewicz A., Jagiello K., Cronin M.T.D. Puzyn T. (2015): Computational modelling - How to overcome problem of insufficient empirical data for nanomaterials? QualityNano Conference, Heraklion, Greece, 15-17 July 2015 (poster presentation).
• Puzyn T. (2015): Towards developing computational models for risk assessment of engineered nanparticles: NANOPUZZLES project summary. 12th International Conference on Nanosciences (NN15). Thessaloniki, Greece, 7-10 July 2015 (oral presentation).
• Gajewicz A., Mikolajczyk A., Zalewska A., Puzyn T. (2015): Safe-by-designe – How computational methods can improve the efficiency of product designing and manufacturing. 12th International Conference on Nanosciences (NN15). Thessaloniki, Greece, 7-10 July 2015 (poster presentation).
• Manganelli S., Leone C., Benfenati E. (2015): QSAR model for cytotoxicity of silica nanoparticles on human embryonic kidney cells. 12th International Conference on Nanosciences (NN15). Thessaloniki, Greece, 7-11 July 2015 (poster presentation).
• Marchese Robinson R.L. Cassano A., Richarz A.N. Cronin M.T.D. (2015): Harvesting data from the nanotoxicology literature to support computational predictions of nanomaterial hazard. 12th International Conference on Nanosciences (NN15). Thessaloniki, Greece, 7-10 July 2015 (poster presentation).
• Gromelski M., Lewandowska W., Gajewicz A., Puzyn T. (2015): Ilościowe modelowanie zależności pomiędzy strukturą modyfikowanych powierzchniowo superparamagnetycznych nanotlenków żelaza (SPIONs), a wychwytem komórkowym w ludzkich komórkach raka trzustki (PaCa2) [in polish]. Quantitative structure-activity relationship between the superparamagnetic iron oxide nanoparticles and the cellular uptake in human pancreatic carcinoma cells (PaCa2). IV Ogólnopolska Konferencja Studentów i Doktorantów Nauk Ścisłych “Człowiek Nauka Środowisko”, Gdańsk, Poland, 25-26 June 2015 (oral presentation).
• Lewandowska W., Gromelski M., Gajewicz A., Puzyn T. (2015): Modelowanie zależności pomiędzy strukturą chemiczną superparamagnetycznych nanotlenków żelaza, a wychwytem komórkowym ludzkich komórek śródbłonka żyły pępowinowej. IV Ogólnopolska Konferencja Studentów i Doktorantów Nauk Ścisłych “Człowiek Nauka Środowisko”, Gdańsk, Poland, 25-26 June 2015 (oral presentation).
• Lewandowska W., Gromelski M., Gajewicz A., Puzyn T. (2015): Modelowanie zależności wychwytu komórkowego superparamagnetycznych nanotlenków żelaza (SPIONs) względem linii komórkowej ludzkich komórek śródbłonka żyły pępowinowej (HUVEC). IV Ogólnopolska Konferencja Studentów i Doktorantów Nauk Ścisłych “Człowiek Nauka Środowisko”, Gdańsk, Poland, 25-26 June 2015 (poster presentation).
• Puzyn T., Gajewicz A. (2015): QSAR and read-acorss for nanoparticles. 8th International Symposium on Computational Methods in Toxicology and Pharmacology Integrating Internet Resources (CMTPI-2015). Chios, Greece, 21-25 June 2015 (oral presentation).
• Mikolajczyk, A., Cybula, A., Gajewicz, A., Zaleska, A., Hirano, S., Puzyn T. (2015): Development of photocatalyst based on Au/Pd@TiO2 nanoparticles by nano-QSPR and safe-by-design approaches, 7th EuroNanoForum Conference, Riga, Latvia, 10-12 June 2015 (poster presentation).
• Gromelski M., Lewandowska W., Gajewicz A., Puzyn T. (2015): Computational approach of modeling statistically significant nano-quantitative structure-activity relationships (nano-QSAR) between the structure of organic surface modifiers in superparamagnetic iron oxide nanoparticles (SPIONs) and the cellular uptake in human pancreatic carcinoma (PaCa2). Chemistry Environment Nanotechnology – International Science Conference (CEN ISC), Gdańsk, Poland, 15-17 April 2015 (poster presentation).
• Lewandowska W., Gromelski M., Gajewicz A., Puzyn T. (2015): Computational approach of modeling statistically significant nano-quantitative structure-activity relationships (nano-QSAR) between the structure of organic surface modifiers in superparamagnetic iron oxide nanoparticles (SPIONs) and the cellular uptake in human pancreatic carcinoma (PaCa2). Chemistry Environment Nanotechnology – International Science Conference (CEN ISC), Gdańsk, Poland, 15-17 April 2015 (poster presentation).
• Sizochenko N., Rasulev B., Puzyn T., Leszczynski J. (2015): Genotoxicity of metal oxide nanoparticles: Classification, Clustering and Causality. XII Annual Conference 14 March 2015 (poster presentation).
• Mikolajczyk A., Rasulev B., Pinto H., Gajewicz A., Leszczynski J., Puzyn T. (2014): Structure and energetics of anatase TiO2 (101) surface-suported Au8 clusters. 11th International Symposium on Recent Advances in Environmental Health Research, 13th International Symposium on Metal Ions in Biology & Medicine, Jackson, MS, USA, 13-16 September 2014 (poster presentation).
• Jagiello K., Avramopoulos A., Chomicz B., Gajewicz A., Papadopouls M., Puzyn T. (2014): The influence of the size on the eletronic properties of the nanometer-sized metal oxides. 14th Southern School on Computational Chemistry and Materials Science, Jackson, USA, 24-25 July 2014 (poster presentation).
• Golbamaki B.N. (2014): Descriptors for nanomaterials. 16th International Workshop on Quantitative Structure-Activity Relationships in Environmental and Health Sciences (QSAR2014), Milan, Italy, 16-20 June 2014 (oral presentation).
• Golbamaki B.N. (2014): Modelling of Nano Metal Oxides’ Genotoxicity. 16th International Workshop on Quantitative Structure-Activity Relationships in Environmental and Health Sciences (QSAR2014), Milan, Italy, 16-20 June 2014 (poster presentation).
• Marchese Robinson R.L. Richarz A.N. Cronin M.T.D. Puzyn T., Benfenati E., Golbamaki B.N. Papadopoulos M.G. Cassano A. (2014): Developing Data Collections for (Q)SAR Modelling of Nanomaterials. 16th International Workshop on Quantitative Structure-Activity Relationships in Environmental and Health Sciences (QSAR2014), Milan, Italy, 16-20 June 2014 (poster presentation).
• Toropova A.P. Toropov A.A. (2014): Optimal descriptor as a translator of eclectic data into models for mutagenicity of fullerene in different conditions. 16th International Workshop on Quantitative Structure-Activity Relationships in Environmental and Health Sciences (QSAR2014), Milan, Italy, 16-20 June 2014 (poster presentation).
• Marchese Robinson R.L. Richarz A.N. Cronin M.T.D. Puzyn T., Benfenati E., Golbamaki B.N. Papadopoulos M.G. (2014): The challenges associated with developing data collections to support modelling of nanomaterial effects. 7th International Nanotoxicology Congress NANOTOX 2014, Antalya, Turkey, 23-26 April 2014 (poster presentation).
• Leonis G., Avramopoulos A., Benfenati E., Cronin M.T.D. Puzyn T., Papadopoulos M.G. (2014): Binding modes and interactions in Human Serum Albumin complexes with Fullerene derivatives. 7th International Nanotoxicology Congress NANOTOX 2014, Antalya, Turkey, 23-26 April 2014 (poster presentation).
• Jagiello K., Chomicz B., Gajewicz A., Papadopoulos M., Puzyn T. (2014): Influence of the size on the physical-chemical properties of NPs. 7th International Nanotoxicology Congress NANOTOX 2014, Antalya, Turkey, 23-26 April 2014 (poster presentation).
• Toropov A.A. (2014): Optimal descriptor as a translator of eclectic data into the prediction of behaviour of complex systems. XXXIV ELBA NANOFORUM, Nanomedicine Workshop, at Laboratories of Biophysics and Nanobiotechnology, DIMES, Genova, Italy, 26 February 2014 (oral presentation).
• Marchese Robinson R.L. Cronin M.T.D. Gajewicz A., Golbamaki B.N. Lubiński Ł., Leszczynski J., Mokshina E., Przybylak K.R. Richarz A.N. Urbaszek P., Puzyn T. (2013): Collection of toxicity, physicochemical and characterisation data to enable modelling of nanomaterial effects. Nanosafety 2013, Saarbrücken, Germany, 20-22 November 2013 (poster presentation).
• Benfenati E. (2013): Metodologie in silico per i nano materiali. Workshop Ambiente e salute: dagli effetti di particolato atmosferico e nanoparticelle alle emissioni di gas serra, Nanomaterials, Milano, Italy, 25 October 2013 (oral presentation).
• Puzyn T. (2013): Towards computational designing of safe nanomaterials. Bioinnovation and ScanBalt Forum 2013, Gdańsk, Poland, 16-18 October 2013 (oral presentation).
• Puzyn T., Gajewicz A. (2013): NANOPUZZLES Project: Modelling properties, toxicity and environmental behaviour of engineered nanoparticles. Nano and Advanced Materials Workshop and Fair NAMF 2013, Warsaw, Poland, 16-18 September 2013 (oral presentation).
• Golbamaki B. N. (2013): Genotoxicity of Metal Oxide Nanoparticles: A New Predictive (Q)SAR Model. EMGS 44th Annual Meeting, Monterey, September 2013 (poster presentation).
• Richarz A.N. Cronin M.T.D. Madden J.C. Lubiński L., Mokshina E., Urbaszek P., Kuz'min V.E. Puzyn T. (2013): Development of computational models for the prediction of the toxicity of nanomaterials. 29th Annual Conference of the Society of Minerals and Trace Elements, Berlin, Germany, 14-15 September 2013 (poster presentation).
• Richarz A.N. Cronin M.T.D. Madden J.C. Lubiński L., Mokshina E., Urbaszek P., Puzyn T., Kuz'min V.E. (2013): Toxicity of nanomaterials: Availability and suitability of data for the development of in silico models. EUROTOX 2013, Interlaken, Switzerland, 1-4 September 2013 (poster presentation).
• Richarz A.N. Cronin M.T.D. Madden J.C. Lubiński L., Mokshina E., Urbaszek P., Puzyn T., Kuz'min V.E. (2013): Database creation, data quality assessment and QSAR models for the toxicity of nanoparticles. MACC-5 – Methods and Applications of Computational Chemistry, Fifth Symposium, Kharkiv, Ukraine, 1-5 July 2013 (poster presentation).
• Mokshina E. Richarz A., Cronin M.T.D. Kuz'min V.E. (2013): NanoQSAR: metal oxides nanoparticles toxicity assessment. MACC-5 – Methods and Applications of Computational Chemistry, Fifth Symposium, Kharkiv, Ukraine, 1-5 July 2013 (poster presentation).
• Toropov A. A. (2013): Modelling toxicity behaviour of engineered nanoparticles. Harmonisation meeting with representatives of other modelling projects consortia of the NMP.2012.1.3-2 Programme, Brussels, 4-6 June 2013 (oral presentation).
• Toropov A. A. (2013): Optimal descriptors as a tool for QSPR/QSAR analyses of substances with complex molecular architecture. Fourth nanosafety annual school ‘Understanding Human Health Effects and Environmental Impacts of Engineered Nanomaterials’, Venice, Italy, 13 March 2013 (oral presentation).

List of Websites:
The project web-site www.nanopuzzles.eu is intended to facilitate dissemination of the information on the Project activities and results, both to project partners and other interested stakeholders. It contains information such as: objectives of each task, direct results, publications links, promotional materials etc.

An initial version of the NanoPUZZLES project website was designed, provisioned and uploaded on the internet (www.nanopuzzles.eu) during the first reporting period and during the second part of the project implementation, the website was being continuously updated, according to the progress of work:
1) the new tab was added called ‘Results and Products’ (http://www.nanopuzzles.eu/results-products) where two sections were created:
a) containing the CORALSEE software for calculating nanodescriptors,
b) containing links to the products related to ISA-TAB-Nano datasets.
2) the tab called ‘Publications and Events’ (http://www.nanopuzzles.eu/publications-events) was updated according to the most recent state of work, in the following four sections:
a) Conference presentations
b) Scientific papers – published
c) NanoPUZZLES Stakeholder Event
d) Final Conference of the NanoPUZZLES project.