Progressive Docking - Deep Learning Based Approach for Accelerated Virtual Screening.

We have developed a novel, hybrid QSAR-docking approach (called ‘progressive docking’) that can speed up the process of virtual screening by enhancing it with Deep Learning models trained on-the-go on produced docking scores. The developed method can, therefore, predict docking outcome for yet unprocessed molecular entries and hence to progressively remove unfavorable chemical structures from the remaining docking base. This approach provides 50–100X speed increase for the standard docking procedures while retaining >90