Mechanochemical co-crystallization: Insights and predictions

Research output: Contribution to journalArticlepeer-review

7 Citations (SciVal)
50 Downloads (Pure)

Abstract

Mechanochemistry integrates mechanical and chemical phenomena to spawn new modes of reactivity, faster reaction kinetics and the discovery of novel materials, in a solvent-free and environmental manner. This simple, yet efficient technique has become a well-established screening technique to discover pharmaceutical co-crystals - components that incorporate a secondary crystalline structure into the lattice of an active pharmaceutical ingredient (API) to improve its physicochemical properties. Today, predicting the reactivity of solid-state materials under mechanochemical conditions remains a major challenge. Here, we explore various machine learning algorithms and identified XGBoost ideal to accurately predict mechanochemical co-crystallization. The model was trained using 1000 co-crystallization events and 2083 chemical descriptors, revealing fundamental insights about mechanochemistry. The model was implemented to screen secondary crystalline structures against a model API, yielding three new mechanochemically-formed co-crystals. This predictive model will accelerate the discovery of novel pharmaceuticals while its insights aid at developing a more sustainable chemical industry.

Original languageEnglish
Article number107416
JournalComputers and Chemical Engineering
Volume153
Early online date27 Jun 2021
DOIs
Publication statusPublished - Oct 2021

Bibliographical note

Funding Information:
This research was supported by Royal Society-Research Grant RSG∖R1∖180090. Jan Gröls (JC) would like to thank the University of Bath for his Ph.D. studentship and the centre for Doctoral Training in Sustainable Chemical Technologies. The authors gratefully acknowledge the Material and Chemical Characterization Facility (MC2) at the University of Bath for technical support and assistance in this work.

Funding Information:
This research was supported by Royal Society-Research Grant RSG?R1?180090. Jan Gr?ls (JC) would like to thank the University of Bath for his Ph.D. studentship and the centre for Doctoral Training in Sustainable Chemical Technologies. The authors gratefully acknowledge the Material and Chemical Characterization Facility (MC2) at the University of Bath for technical support and assistance in this work. The following items are deposited at the University of Bath's Research Data Archive (Gr?ls and Castro Dominguez, 2021):, 1. The generated in-house data used to train the models (including general datasets, raw PXRD data, description, and outcomes), 2. The machine learning training and prediction codes, B.C.D conceived of the presented idea and designed/directed the project. J.G. developed the ML model, designed, and performed the experiments, and analyzed the data. All authors discussed the results and contributed to the final manuscript.

Publisher Copyright:
© 2021 Elsevier Ltd

Keywords

  • Co-crystallization
  • Machine learning
  • Mechanochemistry
  • Solvent-free manufacturing

ASJC Scopus subject areas

  • General Chemical Engineering
  • Computer Science Applications

Fingerprint

Dive into the research topics of 'Mechanochemical co-crystallization: Insights and predictions'. Together they form a unique fingerprint.

Cite this