Machine learning and classical MD simulation to identify inhibitors against the P37 envelope protein of monkeypox virus

Abstract

Monkeypox virus (MPXV) outbreak is a serious public health concern that requires international attention. P37 of MPXV plays a pivotal role in DNA replication and acts as one of the promising targets for antiviral drug design. In this study, we intent to screen potential analogs of existing FDA approved drugs of MPXV against P37 using state-of-the-art machine learning and computational biophysical techniques. AlphaFold2 guided all-atoms molecular dynamics simulations optimized P37 structure is used for molecular docking and binding free energy calculations. Similar to members of Phospholipase-D family , the predicted P37 structure also adopts a β-α-β-α-β sandwich fold, harbouring strongly conserved HxKxxxxD motif. The binding pocket comprises of Tyr48, Lys86, His115, Lys117, Ser130, Asn132, Trp280, Asn240, His325, Lys327 and Tyr346 forming strong hydrogen bonds and dense hydrophobic contacts with the screened analogs and is surrounded by positively charged patches. Loops connecting the two domains and C-terminal region exhibit high degree of flexibility. In some structural ensembles, the partial disorderness in the C-terminal region is presumed to be due to its low confidence score, acquired during structure prediction. Transition from loop to β-strands (244–254 aa) in P37-Cidofovir and its analog complexes advocates the need for further investigations. MD simulations support the accuracy of the molecular docking results, indicating the potential of analogs as potent binders of P37. Taken together, our results provide preferable understanding of molecular recognition and dynamics of ligand-bound states of P37, offering opportunities for development of new antivirals against MPXV. However, the need of in vitro and in vivo assays for confirmation of these results still persists. Communicated by Ramaswamy H. Sarma