Machine Learning Modeling Predicting Vascular Endothelial Growth Factor Receptor 2 (VEGFR2) Inhibitors Structure-Activity Relationships Using Quantum DFT Descriptors

The vascular endothelial growth factor receptor 2 (VEGFR2) is considered the most important marker for endothelial cell development. In particular, this receptor is directly related to tumor angiogenesis regulation. Therefore, several inhibitors of VEGFR-2 are developed, and many of them are now in clinical trials. For the design of new inhibitors against VEGFR2, the half-maximal inhibitory concentration
$(IC_{50})$
is a core step in pharmacological research. In this work,
$IC_{50}$
for the Vascular Endothelial Growth Factor Receptor 2 was studied, and it was modeled using eleven Machine Learning Algorithms. Thirteen molecular descriptors and fingerprints were employed for the in silico modeling. Hyper-parameter tuning was performed for each Machine Learning Algorithm, which helped in the proper selection of parameter values and resulted in improved classification performance. A total of 6678828 models were evaluated, and the best model obtained was a Decision Tree generated from the three most relevant descriptors derived from Density Functional Theory. The best model achieved an average balanced accuracy of 0.75 for the 5-fold cross-validation.