A classification consensus ensemble multitarget neural network model of the dependence of the anxiolytic activity of chemical compounds on the energy of their docking in 17 biotargets was developed. The training set included compounds thathadalready been tested for anxiolytic activity and were structurally similar to the 15 studied nitrogen-containing heterocyclic chemotypes. Seventeen biotargets relevant to anxiolytic activity were selected, taking into account the possible effect on them of the derivatives of these chemotypes. The generated model consistedof three ensembles of artificial neural networks for predicting three levels of anxiolytic activity, with sevenneural networks in each ensemble. A sensitive analysis of neurons in an ensemble of neural networks for a high level of activity made it possible to identify four biotargets ADRA1B, ADRA2A, AGTR1, and NMDA-Glut, which were the most significant for the manifestation of the anxiolytic effect. For these four key biotargets for 2,3,4,5-tetrahydro-11H-[1,3]diazepino[1,2-a]benzimidazole and [1,2,4]triazolo[3,4-a][2,3]benzodiazepine derivatives, eight monotarget pharmacophores of high anxiolytic activity were built. Superposition of monotarget pharmacophores built two multitarget pharmacophores of high anxiolytic activity, reflecting the universal features of interaction 2,3,4,5-tetrahydro-11H-[1,3]diazepino[1,2-a]benzimidazole and [1,2,4]triazolo[3,4-a][2,3]benzodiazepine derivatives with the most significant biotargets ADRA1B, ADRA2A, AGTR1, and NMDA-Glut.

RAGE signal transduction via the RAGE-NF-κB signaling pathway is one of the mechanisms of inflammatory reactions that cause severe complications in diabetes mellitus. RAGE inhibitors are promising pharmacological compounds that require the development of new predictive models. Based on the methodology of artificial neural networks, consensus ensemble neural network multitarget model has been constructed. This model describes the dependence of the level of the RAGE inhibitory activity on the affinity of compounds for 34 target proteins of the RAGE-NF-κB signal pathway. For this purpose an expanded database of valid three-dimensional models of target proteins of the RAGE-NF-κB signal chain was created on the basis of a previously created database of three-dimensional models of relevant biotargets. Ensemble molecular docking of known RAGE inhibitors from a verified database into the sites of added models of target proteins was performed, and the minimum docking energies for each compound in relation to each target were determined. An extended training set for neural network modeling was formed. Using seven variants of sampling by the method of artificial multilayer perceptron neural networks, three ensembles of classification decision rules were constructed to predict three level of the RAGE-inhibitory activity based on the calculated affinity of compounds for significant target proteins of the RAGE-NF-κB signaling pathway. Using a simple consensus of the second level, the predictive ability of the created model was assessed and its high accuracy and statistical significance were shown. The resultant consensus ensemble neural network multitarget model has been used for virtual screening of new derivatives of different chemical classes. The most promising substances have been synthesized and sent for experimental studies.
Peredacha aktivatsii retseptorov RAGE po signal\'nomu puti RAGE-NF-κB iavliaetsia odnim iz mekhanizmov vozniknoveniia vospalitel\'nykh reaktsiĭ, vyzyvaiushchikh tiazhelye oslozhneniia pri sakharnom diabete. Ingibitory RAGE iavliaiutsia perspektivnymi farmakologicheskimi soedineniiami, chto trebuet razrabotki novykh predskazatel\'nykh modeleĭ. Na osnove metodologii iskusstvennykh neĭronnykh seteĭ postroena konsensusnaia ansamblevaia neĭrosetevaia mul\'titargetnaia model\', opisyvaiushchaia zavisimost\' urovnia RAGE-ingibiruiushcheĭ aktivnosti ot affinnosti soedineniĭ k 34 belkam-misheniam signal\'nogo puti RAGE-NF-κB. Dlia étogo na osnove ranee sozdannoĭ bazy dannykh po trekhmernym modeliam relevantnykh biomisheneĭ byla sformirovana rasshirennaia baza dannykh po validnym trekhmernym modeliam belkov-misheneĭ signal\'noĭ tsepochki RAGE-NF-κB. Vypolnen ansamblevyĭ molekuliarnyĭ doking izvestnykh RAGE-ingibitorov iz verifitsirovannoĭ bazy dannykh v saĭty modeleĭ belkov-misheneĭ, opredeleny minimal\'nye énergii dokinga dlia kazhdogo soedineniia v otnoshenii kazhdoĭ misheni i sformirovana rasshirennaia obuchaiushchaia vyborka dlia neĭrosetevogo modelirovaniia. S ispol\'zovaniem semi variantov obucheniia na osnove iskusstvennykh mnogosloĭnykh pertseptronnykh neĭronnykh seteĭ postroeny tri ansamblia klassifikatsionnykh reshaiushchikh pravil dlia prognoza trekh urovneĭ RAGE-ingibiruiushcheĭ aktivnosti po raschetnoĭ affinnosti soedineniĭ k znachimym belkam-misheniam signal\'nogo puti RAGE-NF-κB. S primeneniem prostogo konsensusa vtorogo urovnia vypolnena otsenka prognosticheskoĭ sposobnosti sozdannoĭ modeli, pokazana ee vysokaia tochnost\' i statisticheskaia znachimost\'. S pomoshch\'iu poluchennoĭ konsensusnoĭ ansamblevoĭ neĭrosetevoĭ mul\'titargetnoĭ modeli proveden virtual\'nyĭ skrining novykh soedineniĭ razlichnykh khimicheskikh klassov. Perspektivnye veshchestva sintezirovany i napravleny na éksperimental\'noe izuchenie.

The present paper deals with prediction of cytotoxic activity of 17-picolyl and 17-picolinylidene androstane derivatives toward androgen receptor negative prostate cancer cell line (PC-3). The prediction was achieved applying artificial neural networks (ANNs) method on the basis of molecular descriptors. The most important descriptors (skin permeability (SP), Madin-Darby canine kidney cell permeability (MDCK) and universal salt solubility factor (S+SF)) were selected by using stepwise selection coupled with partial least squares method. The ANN modelling was carried out in order to obtain reliable models which can facilitate further synthesis of androstane derivatives with high antiproliferative activity toward PC-3 cell line. The modelling procedure resulted in three ANN models with the best statistical performance. The obtained results show that the established ANN models can be applied for required purpose.