Frontiers of molecular crystal structure prediction for pharmaceuticals and functional organic materials

Abstract

The reliability of organic molecular crystal structure prediction has improved tremendously in recent years. Crystal structure predictions for small, mostly rigid molecules are quickly becoming routine. Structure predictions for larger, highly flexible molecules are more challenging, but their crystal structures can also now be predicted with increasing rates of success. These advances are ushering in a new era where crystal structure prediction drives the experimental discovery of new solid forms. After briefly discussing the computational methods that enable successful crystal structure prediction, this perspective presents case studies from the literature that demonstrate how state-of-the-art crystal structure prediction can transform how scientists approach problems involving the organic solid state. Applications to pharmaceuticals, porous organic materials, photomechanical crystals, organic semi-conductors, and nuclear magnetic resonance crystallography are included. Finally, efforts to improve our understanding of which predicted crystal structures can actually be produced experimentally and other outstanding challenges are discussed.

Fig. 1. Organic molecular crystal structure is difficult due to the large search space of potential structures (blue dots) on the 0 K crystal energy landscape which are separated by small lattice energy differences. Moreover, the relative free energies between polymorphs vary as a function of temperature and pressure, and not all thermodynamically feasible crystal structures can be readily crystallized experimentally.

Fig. 2. A typical hierarchical crystal structure prediction approach might (1) generate and rank large numbers of candidate structures with an inexpensive force field, (2) refine many of the most promising structures with some method of intermediate accuracy and computational cost, (3) perform dispersion-corrected DFT refinement on a few hundred structures, and (4) perhaps end with free energy calculations on a small number of structures.

Fig. 3. Some CSP procedures involve iterative cycles of force field fitting, structure prediction, and DFT-D structure refinement until the force field and DFT-D crystal energy landscapes are suitably consistent. Adapted with permission from ref. . Copyright 2020 American Chemical Society.

Fig. 4. Delocalization error in GGA and hybrid functionals such as PBE and PBE0 leads to over-stabilization of more planar conformations of the ROY molecular relative to those with a dihedral angle closer to 90°, as compared to high-level coupled cluster benchmarks. This impacts the predicted crystal energy landscape, as will be discussed in Section 4.3.

Fig. 5. Free energy calculations on experimentally-known forms A and B of oxabispidine and several other predicted polymorphs find that form A only becomes the most stable form near room temperature. Adapted with permission from ref. . Copyright 2021 American Chemical Society.

Fig. 6. Predicted crystal energy landscapes for (a) rotigotine and (b) galunisertib. Red points indicate experimentally-observed polymorphs. For rotigotine, a pair of static structures was identified for each of forms I and II which correspond to the two possible conformations of the disordered thiophene ring. The structures labeled “form III” for rotigotine and “GM” for galunisertib have not yet been found experimentally.

Fig. 7. Predicted phase-boundary between hydrate and anhydrate forms of three drugs as a function of temperature and relative humidity. They show nearly quantitative agreement between theory (red lines) and experimentally-derived coexistence points.

Fig. 8. After addressing the DFT delocalization error issues, the predicted crystal energy landscape of ROY shows that the lowest-energy polymorphs have already been discovered experimentally (red). Interestingly, the hypothetical rank #15 structure in blue is predicted to become the most stable structure near 10 GPa. Figure adapted from ref. .

Fig. 9. Crystal energy landscape for the molecule T2, color coded by the methane storage capacity. The α–δ polymorphs represent particularly stable porous crystals that have been crystallized experimentally. T2-γ set a record for the highest surface area among porous organic molecular crystals. Adapted with permission from ref. . Copyright 2017 Springer Nature.

Fig. 10. (a) CSP for the 9-methyl anthracene monomer reactant (blue) and photodimer product (red). The experimental monomer (solid blue) photochemically transforms to the solid-state reacted dimer (SSRD) polymorph, which differs from the solution-grown dimer (SGD) polymorph obtained from solution-phase crystallization. (b) Scheme showing the solid-state topochemical [4 + 4] photodimerization. (c) The photomechanical engine cycle involves fast photodimerization/dissociation reactions (Steps 1 & 3) followed by structural relaxations (Steps 2 & 4) which perform work. Figure adapted from ref. .

Fig. 11. Predicted electron mobilities for various organic semi-conducting material candidates discovered through an evolutionary search. Red dots indicate the mobility for the most stable predicted structure of each species, while blue dots/error bars indicate the landscape-averaged mean mobility and standard deviation for the ensemble of low-energy structures. Three molecules with the highest landscape-averaged mobilities are shown.

Fig. 12. Overlay of the ampicillin crystal structures determined from X-ray diffraction experiments (blue) and the direct NMR crystallography solution (red) which employed a combination of DFTB energies and ShiftML chemical shifts.

Fig. 13. Elimination of (a) lattice energy minima that are labile or effectively equivalent at (b) finite temperatures using molecular dynamics, clustering, and enhanced sampling techniques reduces the number of predicted ibuprofen crystal structures by 65%. Figure adapted from ref. .

Gregory J. O. Beran