Virtual Screening of Multi-Target Agents by Combinatorial Machine Learning Methods

Abstract

Multi-target drugs have greatly attracted the attention and interest in drug discovery. As a joint effort, the Kinetics database of biomolecular interactions and the Therapeutic targets database were upgraded. They can offer informative data in multi-target drug discovery. I explored combinatorial support vector machines (COMBI-SVM) tool for virtual screening of multi-target agents. After the preliminarily tests of COMBI-SVMs for 4 dual-kinase inhibitors pairs (EGFR-Src, EGFR-FGFR, VEGFR-Lck, Src-Lck), I applied the COMBI-SVMs to the identification of dual-target antidepressant agents of 7 target combinations (serotonin transporter paired with noradrenaline transporter, H3 receptor, 5-HT1A receptor, 5-HT1B receptor, 5-HT2C receptor, Melanocortin 4 receptor and Neurokinin 1 receptor respectively). COMBI-SVMs were compared to other VS methods in varies testing sets (e.g. MDDR and PubChem databases). They showed comparable dual-inhibitor yields, moderate to good target selectivity in misidentifying individual-target inhibitors of the same target pair and inhibitors of the other target pairs as dual-inhibitors, low dual-inhibitor false-hit rates in screening large databases MDDR and PubChem.