Comprehensive Assessment of Torsional Strain in Crystal Structures

bucketlist 62

In optimising the potency of a small molecule drug lead, a key objective is to refine its solution-phase conformational landscape so the desired binding pose is pre-conditioned and energetically feasible. However, calculating the conformational global minimum of a drug-like molecule is challenging for both ab initio and molecular mechanics (MM) methods; for example, density functional theory (DFT) is too computationally demanding to achieve extensive sampling of drug-sized molecules and MM force fields have limited accuracy. To tackle these challenges, this 2019 study focused on dihedral angle energetics to assess conformational strain in small molecule binding poses and used a combined quantum mechanics (QM)/MM approach with cloud computing to achieve accurate yet extensive torsional sampling. In order to reduce the degrees of freedom in each simulation, drug-like molecules were fragmented into “minimal torsion fragments” (fragments that encompass the dihedral angle of interest, with a 1 atom environment plus any atoms that prevent the breaking of rings or functional groups). The fragments were subject to gas-phase torsional scans by generating MM profiles for the dihedral angle of interest and optimising the fragment geometry at windows along the profile using DFT. The resulting QM energy profiles were compared with ligand crystal structures from the Cambridge structure database (CSD) and the protein data bank (PDB), revealing that the preferred crystal poses were associated with low-strain dihedral angles. As might be expected, the small fraction of poses that resulted in high torsional strain could mainly be attributed to external steric or hydrogen bonding effects. Despite this, these results suggest that intrinsic torsional energetics are overwhelmingly responsible for the preferred crystal structure poses of small molecules and ligands. With this in mind, the authors suggest a useful workflow for the estimation of conformational strain energy in drug-sized molecules: 1) extract the minimal torsion fragment for each rotatable bond of the input molecule, 2) perform torsional scanning to generate dihedral angle energetic profiles for each fragment, 3) estimate the strain energy of each rotatable bond by mapping the input conformation onto the corresponding energy profile, 4) calculate the total conformational strain energy as a linear sum of the individual strains.

J. Chem. Inf. Model. 2019, 59, 10, 4195–4208