Bayesian Model for the Classification of GPCR Agonists and Antagonists

Abstract

G-protein coupled receptors (GPCRs) are involved in a wide variety of physiological processes and are known to be targets for nearly 50% of drugs. The various functions of GPCRs are affected by their cognate ligands which are mainly classified as agonists and antagonists. The purpose of this study is to develop a Bayesian classification model, that can predict a compound as either human GPCR agonist or antagonist. Total 6627 compounds experimentally determined as either GPCR agonists or antagonists covering all the classes of GPCRs were gathered to comprise the dataset. This model distinguishes GPCR agonists from GPCR antagonists by using chemical fingerprint, FCFP_6. The model revealed distinctive structural characteristics between agonistic and antagonistic compounds: in general, 1) GPCR agonists were flexible and had aliphatic amines, and 2) GPCR antagonists had planar groups and aromatic amines. This model showed very good discriminative ability in general, with pretty good discriminant statistics for the training set (accuracy: 90.1%) and a good predictive ability for the test set (accuracy: 89.2%). Also, receiver operating characteristic (ROC) plot showed the area under the curve (AUC) to be 0.957, and Matthew's Correlation Coefficient (MCC) value was 0.803. The quality of our model suggests that it could aid to classify the compounds as either GPCR agonists or antagonists, especially in the early stages of the drug discovery process.

Comprehensive strategies of machine-learning-based quantitative structure-activity relationship models

“…QSAR is an empirical mathematical model in which regression ( Doo Ho Cho and Bum Tae Kim, 2001 ) and classification ( Choi et al., 2010 ; Choi et al., 2009 ; Kim et al., 2006a , 2006b ; Kim et al., 2008 ; Lee et al., 2017 ; You et al., 2015 ) is performed on many structure-property data to reveal statistically significant correlations between chemical structures and biological properties ( Figure 2 ). A QSAR model can thus predict a new chemical's biological/toxicological properties based solely on the chemical's structure, i.e., without resorting to the time-consuming molecular docking, which makes both the training and application of QSAR highly efficient.…”