Use of machine learning tools and and Structure-Based In Silico Modeling for the search of novel innexin 2 inhibitory molecules as a mitigation method for Caligus rogercresseyi

This research aims to identify selective channel blockers targeting Innexin-2 in the ectoparasite Caligus rogercresseyi to mitigate aquaculture infestations¹. We developed an integrated computational workflow combining support vector machine classification models with atomistic molecular dynamics simulations. A machine learning model evaluated chemical descriptors for flexibility, polar surface area, and charge distribution to screen potential candidate compounds. Structural modeling identified a highly conserved, pore-adjacent intersubunit cavity at the transmembrane interfaces of the octameric channel. Subsequent docking and 1-uS multi-site simulations confirmed that therapeutic modulators (mefloquine, boldine) stably occupy this transmembrane pocket. Mefloquine coordinates with Asp3, while boldine interacts with Phe114 and Ser38. Multi-site simulations successfully evicted false-positive decoys. Mechanistically, an in silico alanine mutation at Asp3 destabilized boldine binding. Electrostatic mapping demonstrated that these inhibitors induce a functional pore-lining charge inversion from negative to positive, blocking ion flow without causing structural pore collapse.