Navigating Solid-Form Screening Using In Silico Methods Validated with Experimental Data for a Drug-like Molecule

Abstract

We investigated the polymorphism of Ziritaxestat (ZTS) by combining sophisticated computational prediction models with experimental crystallization techniques. In order to forecast a stable amorphous state, we developed an improved neural network model. We performed conformational energy calculations using a potential energy scan (PES) and COSMO-RS-predicted activity coefficients to identify a low-energy conformer that could be experimentally obtained in a stable anhydrous form. The predictions of solubility trends in various solvents using COSMO-RS were consistent with the experimental solubility. Using the COSMO-RS function, the solvate probability was predicted. Additionally, the COSMOtherm contact probability calculations predicted the solvation site, while the COSMO BP-TZVPD-FINE level theory determined the hydrogen bonding energy for the solvates and hydrates. We further obtained the hydrate and solvate systems through experimentation. We validated these in silico methods, further approving the proof of concept. With our diverse methodology, we were able to create a nonsolvated crystal form, four different hydrate polymorphs, and various solvates. All novel forms of ZTS (Form A, Form B, Form C, Form D, Form E, Form F, and amorphous) were thoroughly characterized by PXRD, DSC, TGA, NMR, and DVS techniques, among others. The close agreement between calculated conformations and experimental data was proven by single-crystal structure analysis. In this work, we validated the computationally designed algorithms with experimental results to efficiently search for anhydrous, solvates, hydrates, solvate-hydrate, and amorphous forms. This integrated synergistic computational–experimental approach resulted in maximum polymorph search and minimized the polymorphic risk.