Kinetic Exponent Network Analysis

Informations projet

KENA

N° de convention de subvention: 101155395

DOI

10.3030/101155395

Projet clôturé

Date de signature de la CE 22 Mars 2024

Date de début 1 Mai 2024

Date de fin 30 Avril 2026

Financé au titre de

• Marie Skłodowska-Curie Actions (MSCA)

Coût total

Aucune donnée

Contribution de l’UE

€ 181 152,96

Coordonné par

FUNDACION IMDEA NANOCIENCIA
Spain

Periodic Reporting for period 1 - KENA (Kinetic Exponent Network Analysis)

Période du rapport: 2024-05-01 au 2026-04-30

Résumé du contexte et des objectifs généraux du projet

Historically, elucidating molecular structures was a slow, time-consuming step in chemical research. The advent of techniques like NMR dramatically speed up molecular elucidations, allowing chemistry to pursue wider ambitions. Today, the elucidation of chemial reaction networks (CRNs) - i.e. all species, reactions, aggregates - is similarly a strongly time-consuming step. KENA - Kinetic Exponent Network Analysis - seeks to provide techniques that similarly accelerate CRN elucidation, allowing chemistry to pursue wider ambitions.

Europe is trying to recover and re-shore the synthesis of active pharmaceutical ingredients (API’s) that are currently produced 80% outside of the EU. This is required to ensure the access of EU citizens to safe drugs and medication. in order to re-shore API’s in a sustainable and economically viable way, greener and more efficient processes are needed. Two promising approaches are: #1) traditional 1-chemical reaction per reactor needs to change to multi-step cascade synthesis with as many consecutive steps as possible in one go 2 , and #2) batch reactors need to give way for smaller foot-print continuous flow systems. Point #1 means that multiple reactants and product will co-exist in the mixture, thus leading to chemical reaction networks (CRN’s) that are difficult to control. Point #2 allows CRN output to be studied quickly in terms of yield and selectivity (using high-throughput screening), as well as scale-up by parallelization to be implemented.

In recent years, massive improvements have been made due to the advent of cheminformatics and artificial intelligence in exploring and optimizing CRN output, as exemplified by the 2017 acquisition of a Northwestern University start-up by Merck to secure synthetic pathway tool ‘Chematica’ (today rebranded to Synthia TM ) , and recent investments in Postera.ai by Pfizer.) However, for these digital tools to work, the CRN has to be known already, done by mining the chemistry literature and patent space, or by using virtual libraries of compounds that reasonably could be made (like Zinc-15). There is currently no fast method to determine the structure and topology of a CRN using experimental techniques, similar to how chemists determine the structure of molecules using Nuclear Magnetic Resonance and other analytical methods.

The Kinetic Exponent Network Analysis (KENA) project aims to provide such a tool, to extract CRN topology directly from the time-dependence of experimental variables in a power law expansion.


Objectives.
1) KENA theory: establish mathematical foundations of KENA and its behavior in a wide variety of chemical contexts.
2) Decompositions: algorithms to split up spectral data in power series contributions.
3) Flow reactor platform: flow and microfluidic platforms tailored to KENA analysis. Experimental validation.
4) Improvement through additional information: expand the KENA concept to incorporate more index-like information
from measurements to further narrow down candidate structures. Relate network structure and properties to measurable indices.

Travail effectué depuis le début du projet jusqu’à la fin de la période considérée dans le rapport et principaux résultats atteints jusqu’à présent

WP1 Kena Theory
1.1. A rigorous general construction for KENA theory was realized using (Taylor) series expansion arguments for dynamical systems on a hypergraph structure,
1.2 This enabled:
- exact computation of coefficients in the KENA series
- the study of the role of experimental resolution, resolvability of species, data dimension, hidden species
- a universal mechanism of low-dimensional response for arbitrarily large networks
1.3 The exponential filtering power of indices and role of the number of experiments was demonstrated through case studies.
1.4 A methodology to use KENA for data with sudden discontinuities (e.g. phase transitions) was formulated
1.5 KENA Theory was validated on experimental data from the literature. A KENA-derived data dimension analogue was in turn used to obtain similar results, and published as a tutorial with case studies.
A paper on the full KENA theory was drafted.

WP2 KENA Spectral decompositions
2.1 Natural spectral decompositions were directly derived using results from 1.1.
2.2 Numerical schemes were implemented for decompositions.
2.3 And applied to real and synthetic UV-VIS data
2.2-2.3 revealed that agnostic extraction of a decomposition from data is nontrivial: different power series can fit the same experimental data remarkably well, especially with added noise.
This limits the interpretative power of the decomposition, and additional constraints need to be derived to enhance this procedure, which will be pursued in followup work.

WP3 Flow reactor implementations
3.1-3.2 4 syringe pumps were built and a variety of reactors were designed, 3d printed and optimized. A simple portable setup was realized as well as a glass chip UV-VIS setup.
It was concluded that more broadly applicable - but more expensive - techniques (IR, RAMAN, MS) need to be pursued to get the most out of KENA. (followup)
3.3 An ambitious ongoing collaboration to study the Formose reaction using new data analysis tools was started, which has confirmed our new hypotheses on the nature of this reaction.
New data analysis methodologies were applied to systems in the host lab, including two papers (in review) on supramolecular assembly and a chemical oscillator (in preparation).

WP4 Expanding KENA
WP4 involved characterizing how notions of dimension related to kinetic exponents, how these relate to CRN structure, and to experimental data, and was assessed in four instances.
4.a An approach for linear networks was developed that retains and eliminates candidate reactions through successive datasets (experiments),
4.b a mysterious oscillator in the host group with unidentified network was studied. A theory relaing oscillatory networks to their structure was developed and published.
This theory enabled the identification and assesment of key candidate hypotheses of the experiments (in preparation).
4.c discoveries made in 3.3 on supramolecular chemistry could be combined with mathematical insights from the structural theory in 4.b which led to the formulation of a
a theory of copolymerization and laws relating polymer sequences to reaction mechanisms.
4.d Extensions of the KENA concept were formulated that employ on-device protocols to produce higher-dimensional data and validated against synthetic data.
Funding was secured to pursue this new KENA concept.

Main Achievements:
10 articles prepared, among which 2 accepted, 4 preprinted, 3 in review.

KENA delivered several mature new theories, which each have been directly applied to experimental data, solved open questions and helped identify new questions
- Data Dimension theory
- Kinetic Exponent Theory
- Dimensional Theory of Phases and Psuedophases
- Theory of Oscillating Reactions in Reaction Networks via Parameter-rich Kinetics

Progrès au-delà de l’état des connaissances et impact potentiel prévu (y compris l’impact socio-économique et les conséquences sociétales plus larges du projet jusqu’à présent)

Kena delivered tested theories, which make predictions that go beyond the state of the art and thereby allowed to resolve open questions.
In particular:
- Data Dimension theory
- Kinetic Exponent Theory
- Dimensional Theory of Phases and Psuedophases
- Theory of Oscillating Reactions in Reaction Networks via Parameter-rich Kinetics

For the interfence of reaction networks, the theories developed in KENA are unique in being derived from a chemistry-wide perspective and aimed at the general resolution of the problem by combining enough interpretable indices to systematically filter candidate networks.
This stands in contrast with methodologies that combine a small number of clues to generate a small number of candidate hypotheses, which are in turn assessed by fitting, which cannot be extended beyond small networks.
As such, KENA - and more broadly, its broader filtering philosophy - is at present the only scaleable approach for this challenge.

An extension of the KENA concept was developed theoretically using on-device protocols (e.g. a photochemical perturbation).
An initial validation of the theoretical concept was validated using synthetic data.
The concept will be pursued further in the next 5 years.