Prediction of Cytotoxicity Against HepG2 by Quantitative Structure-Activity Relation (QSAR) Modelling

Hepatocellular carcinoma (HCC) is the dominant subtype of liver cancer with very low survival rate but chemotherapy for HCC still in grey zone due to limited efficacy and high toxicity profile approved drugs raising heavy demand on drug development HCC. The study aimed establish a desirability based quantitative structure activity relation (QSAR) model predict chemical compounds against one cell line (HepG2). Different support vector machine (SVM) models were constructed ensembled 10 virtual screening protocols. These protocols validated by an external dataset combination decoys as interference. Results showed that ensemble exhibited improved area under Receiver Operating Characteristic Curve (ROC), sensitivity, specificity compared base training test set. When being recover known active molecules mixture inactive decoy compounds, all have good performance Boltzmann-Enhanced Discrimination ROC (BEDROC) enrichment factor (EF). best protocol BEDROC 0.63 EF 29.55 was suitable further HepG2 line. HIGHLIGHTS Ensemble structure-activity relationship single Cytotoxicity prediction transformed score general output easily integration multi-objective optimization Virtual cytotoxicity well GRAPHICAL ABSTRACT