TrixX-BMI: Fast virtual screening using compressed Bitmap Index Technology for efficient prefiltering of Compound Libraries
Jochen Schlosser and Matthias Rarey
Due to the enormous size of available compound libraries the development of extremely fast structure-based virtual screening tools is an important aim for chemogenomics and drug discovery. Most molecular docking tools process a given library by docking each ligand iteratively into the binding site. These tools perform well concerning accuracy but their runtime is linear in the number of compounds to be screened. Our new approach for high-throughput virtual screening tries to break this barrier by reducing the number of candidate ligands for a given receptor via a descriptor based filtering procedure using fast bitmap indices. These indices holding compound descriptors based on triangle geometry, chemical interactions and shape are queried using similar descriptor triangles from the active site. Because most conformational issues are preprocessed and are part of the descriptor most query matches yield a reasonable ligand pose. Results show that TrixX-BMI is an efficient tool for virtual screening with average runtime in sub-second range with comparable enrichment rates.