A MACHINE LEARNING-BASED QUANTITATIVE STRUCTURE- ACTIVITY RELATIONSHIP STUDY FOR THE CARCINOGENIC ACTIVITY OF PHENYLETHYLAMINE
Jamshid Kayumov, Usmanov, Durbek, Rasulev, Bakhtiyor · Zenodo (CERN European Organization for Nuclear Research) · 2022
ABSTRACT: The psychotomimetic activity of substituted phenethylamines as psychedelic drugs is predicted using a novel model. The structure-activity study was carried out using a quantitative structure-activity relationship (QSAR) method, which took into account the molecular structures and activities of the intended phenethylamine derivatives. 118 different substituted phenethylamines with published psychotomimetic activity values are used in this research. The QSAR analysis was carried out using a hybrid approach that included a genetic algorithm for variable selection and multiple linear regression analysis. A quantum-chemical analysis using a semi-empirical approach was used to find a stable conformation and generate additional descriptors for the QSAR study. As a result, a number of models were developed, with the best predictive performance coming from a ten-variable model with r2 = 0.7428 and q2LOO = 0.6738. A leave-one-out technique, external set, and y-scrambling methods were used to test the best model's robustness and predictability. With the external set, the model's predictive ability was proven, with r2ext = 0.7365. The developed model can be used to predict the psychotomimetic activity of newly synthesized and untested organic compounds.