A MACHINE LEARNING-BASED QUANTITATIVE STRUCTURE- ACTIVITY RELATIONSHIP STUDY FOR THE CARCINOGENIC ACTIVITY OF PHENYLETHYLAMINE

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.