Conformer generation with OMEGA: learning from the data set and the analysis of failures

Area of Science:

• Computational chemistry
• Structural biology
• Cheminformatics

Background:

• A high-quality validation dataset for conformer generators was previously established using structures from the Protein Data Bank (PDB) and the Cambridge Structural Database (CSD).
• The performance of the OMEGA conformer generator was tested on these datasets.

Purpose of the Study:

• To assess the suitability of existing validation datasets for testing conformer generators.
• To identify and analyze the failure cases of the OMEGA conformer generator.
• To compare property space coverage between validation and parent compound sets.

Main Methods:

• Comparison of property space coverage between validation and parent compound sets.
• Introduction and application of torsion fingerprinting for dissimilarity assessment.
• Development and utilization of a new metric, RMSTanimoto, for comparing conformations.
• Analysis of OMEGA's failure cases, particularly those related to crystallographic structures.

Main Results:

• The RMSTanimoto metric effectively identifies reproduction failures, especially for smaller molecules.
• Analysis of failures highlights the issue of strain in crystallographic structures.
• Some failure cases present significant challenges for conformer generation engines, indicating areas for improvement.

Conclusions:

• Existing validation datasets may not fully represent the applicable property spaces.
• The RMSTanimoto metric is a valuable tool for identifying computational conformation generation failures.
• Further research is needed to address residual failure cases and improve conformer generation algorithms, with attention to the limitations of crystallographic data.