Year: 2026
Type: preprint
Abstract: This paper presents an efficient conformational filter protocol for handling structures with topological symmetry implemented in the open-source molecular modelling program curcuma. This approach addresses the challenge of comparing molecular conformations in which symmetrically equivalent atoms may have different indices in the structure files. Descriptors that are independent of the storing order of the atoms (energy differences, rotational constants, and Vietoris-Rips (VR) persistence barcodes) are used to efficiently predict potential structural similarities before performing computationally expensive permutations on selected structure pairs using distinct permutation approaches implemented in curcuma or the external tool molalign. The filter protocol operates in three phases: 1) initial RMSD-based filtering and data-set-specific training of the atom index-independent descriptor thresholds, 2) permutation-based filtering using parameterised descriptors, and 3) final filtering using stored permutation vectors. Our evaluation using 395 complex conformations optimised at different levels of theory shows that although conformational geometries typically remain similar across different theoretical methods; their relative energies can vary significantly between different levels of theory. This highlights the importance of efficient conformational filtering in identifying structurally unique conformers, particularly when faster semiempirical quantum chemical methods are used, which might incorrectly evaluate certain conformers as energetically unfavourablein contrast to slower more accurate quantum mechinal methods. A key finding is that performing all possible RMSD calculations does not necessarily yield the optimal filtering result, as the final outcome depends significantly on the specific order of the structure comparisons.
Source: ChemRxiv
Author: Conrad Hübler
Institution: TU Bergakademie Freiberg
Language: English
Subfield: Drug Discovery
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7213871809