Биоминеральные добавки для самозалечивания бетона

Aim. Construction of a neural network multi-target regression model of the dependence of the SERT affinity of chemical compounds on the energies of their docking into relevant biotargets, sensitivity analysis of the resulting model and identification of key biotargets that most significantly affect the affinity of compounds to the serotonin transporter. Materials and Methods. Data on 2D structure and SERT inhibitory activity of 3436 known compounds and drugs; 478 3D models of human target proteins relevant to SERT inhibitory activity; 10 original and licensed computer programs. The following were performed: construction of optimized 3D models of SERT inhibitory compounds and drugs; identification of biotargets relevant to SERT inhibitory activity; finding binding sites and valid 3D models of these biotargets; ensemble docking of known SERT inhibitors to relevant biotargets, calculation of minimum docking energies, formation of a training set; training and testing of regression neural networks, selection of the best neural network; sensitivity analysis of the neural network multi-target model, identification of key biotargets that significantly affect SERT affinity. Results. 21 biotargets relevant to SERT inhibitory activity were identified, 25 binding sites were revealed for them, 68 valid 3D models were found. Ensemble docking of known SERT inhibitors into relevant biotargets was performed, a training set was formed. 7500 neural networks were trained, a regression neural network model of the dependence of SERT affinity of compounds on their affinity to 24 sites of 20 relevant biotargets was built, having a correlation coefficient RTotal=0.885 and statistical significance p<5·10ˉ⁷. A sensitivity analysis of this model was performed, 12 key biotargets were identified that most significantly affect SERT affinity. Conclusions. Using artificial intelligence methods, 12 key biotargets were found that most significantly affect the affinity to SERT, of which six targets are characterized by a positive effect (ADRA1A, AMPA, DRD1, GAT1, HTR2A Alo, HTR3A) and six targets are characterized by a negative effect (HTR2C, JAK2, MAPK11, NMDA Block, ROCK2, SLC18A2). These biotargets reflect the multitarget mechanism of serotonergic regulation of cellular stress in inflammation and various functional disorders.