Virtual Screening Of Cytochrome P450Inhibitors And Molecular Docking

The cytochrome P450(CYP450) enzymes are major drug-metabolizing enzymes,which are responsible for the metabolism of more than90percents of clinical drug’smetabolism. Drug-drug interactions mediated by CYP450may lead to clinicallysignificant adverse drug-drug interactions (DDIs). Therefore, early prediction ofCYP450inhibitors is of both theoretical significance and practical values in drugdiscovery and personalized medicine.In this paper, partial least squares discriminant analysis (PLSDA) and supportvector machine (SVM) were employed to develop virtual screening models for3majorCYP isoforms (1A2,2C9and2C19) based on the molecular hologram and MACCSdescriptors, and achieved good results. In comparison with earlier studies, the optimalSVM screening model established on a large datase is extremely simple, robust,predictive, and interpretable. Then, molecular docking was performed to explore theinteraction modes of CYP inhibitors with key residues of binding pockets of the3CYPisoforms, respectively.The main results are as follows:①Prediction models of CYP2C19inhibitors and molecular docking: Molecularhologram and MACCS descriptors were used to characterize the12,240compoundswith known CYP2C19inhibitory activities. Then PLSDA and SVM were employed toestablish classification models based on molecular hologram and MACCS descriptors.An optimal SVM model based on17molecular holograms and MACCS descriptors wasobtained, of which the Acc, Sen, Spe, and MCC were75.53%,78.13%,73.54%, and51.24%for a training set (n=5,387), and75.29%,76.69%,74.22%, and50.51%for atest set (n=5,383), respectively. The optimal SVM model was further validated by anindependent dataset (n=1,470), of which the Acc, Sen, Spe, and MCC were76.53%,47.12%,81.38%, and23.73%. The results of molecular docking show that themolecular steric complementarity, hydrophobicity, and hydrogen bonding interactionsare closely related to the activities of CYP2C19inhibitors. Meanwhile, the residuesPhe476, Phe114, Ala297, Thr301, and Glu300play important roles in the interactionsbetween the CYP2C19inhibitors and the receptor.②Prediction models of CYP2C9inhibitors and molecular docking: Molecularhologram and MACCS descriptors were used to characterize13,890compounds with known CYP2C9inhibitory activities, and then SVM and PLSDA were used to establishclassification models, respectively. The optimal SVM model based on14molecularholograms and MACCS descriptors was obtained. The Acc, Sen, Spe, and MCC of theSVM model were73.78%,78.07%,71.67%, and47.03%for a training set (n=6179),and72.75%,78.11%,70.12%, and45.51%for a test set (n=6177), respectively. The Acc,Sen, Spe, and MCC for an independent dataset (n=1534) were71.90%,48.78%,74.67%,and16.17%. The results of molecular docking show that hydrophobicity, hydrogenbonding interactions and π-π interactions are closely related to the inhibitory activities.Also, the residues Arg108, Phe476, Phe114and Glu300play important roles in theinteraction between the inhibitors and the CYP2C9receptor.③Prediction models of CYP1A2inhibitors and molecular docking: Based on themolecular holograms and MACCS description method, SVM and PLSDA were used todevelop classification models of CYP1A2inhibitors. The optimal SVM model based on19molecular holograms and MACCS descriptors was obtained, of which the Acc, Sen,Spe, and MCC were80.05%,83.51%,77.09%, and60.43%for a training set (n=6396).The Acc, Sen, Spe, and MCC for two test sets were77.89%,81.74%,74.59%,56.19%(Test set I, n=6395) and61.57%,53.21%,79.93%,31.04%(test set II, n=2581),respectively. The results of molecular docking show that hydrophobicity, π-πinteractions, steric complementarity and molecular planarity are closely related to theinhibitory activities. The results also show that aromatic residues play an important rolein the interaction between the inhibitors and the CYP1A2receptor.④The results of the prediction models of3major CYP isoforms show thatcombination of molecular hologram and MACCS descriptors can enhance predictivepower effectively. The results of molecular docking show that molecular stericcomplementarity and hydrophobic interactions are two key factors influncing inhibitoryactivities aganist3major CYP isoforms. At the same time, the results also show there isno obvious correlationship between docking scores and inhibitory activities,which suggests the scoring function of Surflex is unable to give accurate evaluations onthe activities of CYP inhibitors.