Neural network based on correlation convolution of multi-target energy spectra of multiple docking: a new method of artificial intelligence in the search for pharmacologically active substances
Year: 2024
Type: article
Abstract: 22 biotargets relevant to the anxiolytic activity of chemical compounds have been identified. Multiple docking of known anxiolytic substances in 22 valid 3D models of relevant biotargets has been performed. The energy spectra of multiple docking of these compounds have been calculated for the entire volume of relevant target proteins. Correlation convolution of the energy spectra of multitarget multiple docking in 22 convolution variables has been performed. The one-way ANOVA method has shown high statistical significance of using the obtained convolution variables to estimate the integral multitarget affinity of ligands to a set of relevant biotargets. Neural network models of the dependence of the anxiolytic activity level of chemical compounds on the specified convolution variables have been formed. Very high statistical significance of the constructed neural network models has been shown. As a result, a new artificial intelligence method was developed for in silico search for pharmacologically active substances based on a convolution correlation neural network and energy spectra of multitarget multiple docking.
Authors: Pavel M. Vasiliev, Maxim A. Perfiliev, Arina V. Golubeva, Andrey N. Kochetkov
Institution: Volgograd State Medical University
Language: English
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4406673023