Automatic recognition of ligands in electron density by machine learning

• DOI: 10.1093/bioinformatics/bty626
• Publication Date: 2018-07-14T17:43:43Z
• Authors: Marcin Kowiel; Dariusz Brzezinski; Przemyslaw J Pore...; Ivan G Shabalin; Mariusz Jaskolski; Wladek Minor
• Keywords: Identification; Benchmark (surveying); Protein ligand

The correct identification of ligands in crystal structures protein complexes is the cornerstone structure-guided drug design. However, cognitive bias can sometimes mislead investigators into modeling fictitious compounds without solid support from electron density maps. Ligand be aided by automatic methods, but existing approaches are based on time-consuming iterative fitting.Here we report a new machine learning algorithm called CheckMyBlob that identifies experimental In benchmark tests portfolios up to 219 931 ligand binding sites containing 200 most popular found Protein Data Bank, markedly outperforms methods for identification, some cases doubling recognition rates, while requiring significantly less time. Our work shows improve automation structure and accelerate screening process macromolecule-ligand complexes.Code data available GitHub at https://github.com/dabrze/CheckMyBlob.Supplementary Bioinformatics online.