(698a) The Synergy of Computational Modeling, Machine Learning, and Experiments in Pharmaceutical Solid-State Research and Development

Computational modeling and machine learning (ML) have achieved recognition as vital technologies in drug discovery. Using these technologies, XtalPi has developed a series of methods to guide and enhance experimental workflows for drug development:

- Virtual multicomponent screening to predict the likely salts, cocrystals, and solvates of a compound and recommend solvents for formation experiments. These virtual screenings can be used to reduce the number of wet-lab experiments performed and increase the likelihood of generating the desired result.

- Crystal structure prediction (CSP) predicts all possible polymorphs of a compound and ranks them by thermodynamic stability. The CSP structure-energy landscape can reveal if the most stable polymorph has been discovered and complements experimental XRPD, SC-XRD, and MicroED techniques for crystal structure determination.

- AI-enhanced crystallization (Xtal²) is a machine learning model—constructed from more than 100k virtual and 10k experimental data—used to recommend crystallization strategies based on molecular structure information. Combining Xtal2 with autonomous workstations allows intelligently designed experiments to be run 24x7.

- Morphology prediction calculations reveal how variables like solvent and additives affect the particle shape of a crystallized compound. These calculations reveal if an undesirable morphology (e.g. needles) is expected and can be avoided by crystallization conditions, or if engineering solutions (e.g. milling) will be required.

Keywords: CSP, computational modeling; polymorphs; particle morphology; automation; ML

References

1. Sun, G., et al. 2020. “Virtual Coformer Screening by Crystal Structure Predictions: Crucial Role of Crystallinity in Pharmaceutical Cocrystallization.” J. Phys. Chem. Lett. 11 : 8832-8838. doi: 10.1021/acs.jpclett.0c02371

2. Sekharan, S., et al. 2021. “Selecting a Stable Solid Form of Remdesivir using Microcrystal Electron Diffraction and Crystal Structure Prediction.” RSC Adv. 11 : 17408. doi: 10.1039/D1RA03100G

3. Yuan, J., et al. 2021. “Virtual Coformer Screening by a Combined Machine Learning and Physics-based Approach.” CrystEngComm 23 : 6039-6044. doi: 10.1039/D1CE00587A

4. Gui, Y., et al. 2022. “Crystal Energy Landscape of Nifedipine by Experiment and Computer Prediction.” Cryst. Growth Des. 22(2) : 1365-1370. doi: 10.1021/acs.cgd.1c01317

5. Bellucci, M.A., et al. 2023. “Effect of Polymer Additives on the Crystal Habit of Metformin HCl.” Small Methods. 220 : 2201692. doi: 10.1002/smtd.202201692

6. Abramov, Y.A., et al. 2023. “Cocrystal Synthesis Through Crystal Structure Prediction.” Mol. Pharm. 20(7) : 3380-3392. doi: 10.1021/acs.molpharmaceut.2c01098