Short Nonbonded Contact Distances in Organic Molecules and Their Use as Atom-Clash Criteria in Conformer Validation and Searching

Short, intramolecular nonbonded contact distances from a large sample of organic molecules retrieved from the Cambridge Structural Database (CSD) have been analyzed. With the exception of the element pairs N···S, O···P, O···S, and S···S, the first percentiles of X···Y distance distributions (X, Y = C, Br, Cl, F, N, O, P, S) are well estimated by ∑vdw −0.5 Å, where ∑vdw is the sum of the Bondi van der Waals radii. The 0.1th percentiles are typically a further 0.1 Å shorter. Some 99% of well-refined organic molecules from the CSD have no nonbonded contacts shorter than the 0.1th percentile and no more than two contacts shorter than the first percentile. This can be used as the basis of an atom-clash test for validating less precise crystal-structure data, such as the geometries of protein ligands. In principle, the same test can be used in molecular modeling to identify and filter out unacceptable conformations generated in a conformational search. This is complicated by the fact that conformer generation is usually performed on molecular models with standard bond angles that are not relaxed during the search. In consequence, conformations often appear to contain untenable nonbonded contacts, which would, however, be removed by bond-angle relaxation. This is particularly likely for molecules containing conjugating substituents bonded to adjacent atoms of an aromatic ring, or on the same side of a double bond. Other molecules particularly likely to be affected are those containing rings or other bulky groups separated by a single-atom linkage, and those with the capacity to form intramolecular hydrogen bonds. The problem is greatly ameliorated by the fact that there are many ways to approximate a true conformation, leading to an increased probability that at least one of the approximations will satisfy atom-clash criteria.