Periodic Reporting for period 3 - HarmonicSS (HARMONIzation and integrative analysis of regional, national and international Cohorts on primary Sjögren’s Syndrome (pSS) towards improved stratification, treatment and health policy making)
Período documentado: 2020-01-01 hasta 2020-12-31
HarmonicSS is an EU funded project focused on clinical unmet needs of primary Sjogren’s syndrome (pSS). The main goal of HarmonicSS was to bring together the most important clinical centers specialized in SS across Europe and gather the maximum quantitative and qualitive clinical data in order to address clinically significant questions related to SS. To accomplish such an ambitious vision, distinct scientific fields were engaged and the project consisted of 21 clinical and 13 technical partners. Therefore, HarmonicSS was designed and developed in both clinical and technical level with common denominator the harmonization process which represents the core of the project. The harmonization process aims to match and transform the heterogeneous terminology used by different clinical partners for the various aspects of the disease, into a common set of terms designated as reference model. The common “scientific language” of reference model offers the opportunity to align clinical information related to SS, ensuring an automated process which homogenize and thus “harmonize” clinical data. This initial step is absolutely necessary in order to handle and utilize SS related data in an optimal way, given that the common reference model can be further expanded and enriched, reflecting the “technical” plasticity provided by the infrastructure which hosts the HarmonicSS project. Apart from the harmonization process itself, technical challenges such as data curation, federated analysis as well as legal issues related to GDPR and data sharing, have been achieved within the HarmonicSS project.
The clinical aspects of the project, are summarized as follows: a) define distinct clinical phenotypes of the disease along with the prevalence of significant clinical, laboratory and histologic parameters in a totally harmonized cohort through integration analysis, b) explore potential associations between the histopathology of labial minor salivary gland (LMSG) biopsy and the clinical phenotype of the disease, c) validate old and discover novel biomarkers, d) study and develop lymphoma classification and prediction models, and e) employ data driven approaches contributing to the transition into the era of precision medicine.
The clinical aspects of the project, are summarized as follows: a) define distinct clinical phenotypes of the disease along with the prevalence of significant clinical, laboratory and histologic parameters in a totally harmonized cohort through integration analysis, b) explore potential associations between the histopathology of labial minor salivary gland (LMSG) biopsy and the clinical phenotype of the disease, c) validate old and discover novel biomarkers, d) study and develop lymphoma classification and prediction models, and e) employ data driven approaches contributing to the transition into the era of precision medicine.
Overall, the HarmonicSS project after conducting extensive research at the 3 levels of analysis, successfully addressed the previously described unmet needs of SS: i) the prevalence of major and commonly used in clinical practice clinical features were estimated in 7,551 retrospectively harmonized patients and the effect of gender, early and late disease onset, presence of cryoglobulins, absolute seronegativity, geolocation and heavy inflammatory infiltration within LMSG, on the clinical expression of the disease was studied, ii) the clinical spectrum of specific subsets of pSS patients was extensively investigated including males, cryoglobulinemic and lymphoma patients, early and late disease onset patients and those with high focus score (FS), iii) cryoglobulinemia, total ESSDAI score at SS diagnosis, salivary gland enlargement, rheumatoid factors and male gender were identified as risk factors associated with lymphoma, creating a risk stratification landscape for lymphoproliferative disorders in SS, iv) older biomarkers such as CXCL13 or traditional lymphoma predictors were validated at least to some extent and new biomarkers were discovered including miRNA200b-5p in MSG biopsy specimens and serum tissue lymphopoietin serum protein (TSLP).
Major progress and innovation were achieved from the technical point of view. The data sharing assessment module provides functionalities for the upload of legal and ethical documents, the evaluation of GDPR compliance of these documents and the subsequent application of beyond the state-of-the-art data curator mechanisms to enhance the quality of clinical data in terms of accuracy, relevance and completeness. The data sharing management module handles data access to the private cloud space of each data provider. The cohort data harmonization module provides functionalities for aligning the heterogeneous datasets using ontology-based mechanisms. The data mining services include several tools, which can support both local and federated learning scenarios. These functionalities have been used for clinical scenarios in order to address the clinical unmet needs of pSS. The genetic data analytics services module offers functionalities for mining association rules with pre-defined support and confidence intervals across genetic datasets towards the discovery of associations between clinical sub phenotypes and SNPs. The visual analytics module provides tools for extracting hidden patterns within the cohort data through the implementation of high-performance visualization methods. The social media analytics services module offers a single-point access to pSS-related social media posts and related content with filtering options. The health policies impact assessment services module enables the evaluation of user-defined health policy scenarios by assessing whether health impact and cost of scenarios are positive or not in the existing healthcare systems. The patient selection tool for multinational clinical trials provides functionalities for the targeted selection of patients across the harmonized cohort data for multinational clinical trials, given a specific set of pre-defined criteria. The salivary gland ultrasonography image segmentation (SGUS) module applies deep learning algorithms to distil knowledge from SGUS images towards the automated segmentation of the salivary gland and the classification of SGUS images according to a pre-defined scoring system. Finally, the training tool provides educational material to both non-clinical and clinical experts including text, image or video.
Of high importance were also the results coming from application of health policy and process evaluation. Findings from survey data show variations in access, volumes of treatments delivered to pSS patients and also their perceived quality of life and satisfaction for SS care across Europe.
Major progress and innovation were achieved from the technical point of view. The data sharing assessment module provides functionalities for the upload of legal and ethical documents, the evaluation of GDPR compliance of these documents and the subsequent application of beyond the state-of-the-art data curator mechanisms to enhance the quality of clinical data in terms of accuracy, relevance and completeness. The data sharing management module handles data access to the private cloud space of each data provider. The cohort data harmonization module provides functionalities for aligning the heterogeneous datasets using ontology-based mechanisms. The data mining services include several tools, which can support both local and federated learning scenarios. These functionalities have been used for clinical scenarios in order to address the clinical unmet needs of pSS. The genetic data analytics services module offers functionalities for mining association rules with pre-defined support and confidence intervals across genetic datasets towards the discovery of associations between clinical sub phenotypes and SNPs. The visual analytics module provides tools for extracting hidden patterns within the cohort data through the implementation of high-performance visualization methods. The social media analytics services module offers a single-point access to pSS-related social media posts and related content with filtering options. The health policies impact assessment services module enables the evaluation of user-defined health policy scenarios by assessing whether health impact and cost of scenarios are positive or not in the existing healthcare systems. The patient selection tool for multinational clinical trials provides functionalities for the targeted selection of patients across the harmonized cohort data for multinational clinical trials, given a specific set of pre-defined criteria. The salivary gland ultrasonography image segmentation (SGUS) module applies deep learning algorithms to distil knowledge from SGUS images towards the automated segmentation of the salivary gland and the classification of SGUS images according to a pre-defined scoring system. Finally, the training tool provides educational material to both non-clinical and clinical experts including text, image or video.
Of high importance were also the results coming from application of health policy and process evaluation. Findings from survey data show variations in access, volumes of treatments delivered to pSS patients and also their perceived quality of life and satisfaction for SS care across Europe.
The HarmonicSS project offers clinically important perspectives far beyond the narrow field of pSS, encompassing any complex and systemic disease: 1) The reference model and the harmonization process have been proven fundamental tools to describe any complex systemic disease such as systemic autoimmune diseases or diabetes mellitus, with SS being the prototype example, not only for research purposes but also for future construction of medical files in various health systems, 2) The introduction of novel, hybrid data driven approaches such as the FCBF/ binary multivariable logistic regression for integrated type of analysis, can be utilized to design classification and prediction models for diseases, instead of the typical backward, forward or stepwise logistic regression models used so far, and 3) the advanced and high level of the infrastructure which includes a variety of tools and services along with not only integration but also federated type of analysis, and above all the plasticity to adjust to future needs, may cover a wide spectrum of systemic diseases in terms of research and therefore can be adopted also by other fields.