Modeling the Binding Affinity of p38a MAPKinase Inhibitors by Partial Least SquaresRegression

The p38 mitogen-activated protein kinase is activatedby environmental stress and cytokines andplays a role in transcriptional regulation andinflammatory responses. Factors influencing theactivity and selectivity of the p38a mitogen-activatedprotein kinase inhibitors have been investigatedin this paper by inspecting the bindingorientation and the possible residue-inhibitorinteractions in the binding site. The binding patternof a set of 45 different inhibitors against p38amitogen-activated protein kinase was studiedthrough Molecular Dynamic Simulations of theprotein-inhibitor complexes. Further, Partial LeastSquares regression was used to develop a QuantitativeStructure Activity Relationship model topredict the binding affinities of ligands. Theselected model successfully predicted the test setwith a Root Mean Square Error of Prediction of1.36. The regression coefficients and the VariableImportance in Projection plots highlighted the residue-inhibitor interactions which exhibited thelargest absolute effect on the ligand binding, suchas the van der Waals interaction with LYS50,ILE81, ASP165; electrostatic interactions withSER29, LEU164; hydrogen bonds with MET106;and total energy interaction with SER29 andLEU83.