Prediction Of Protein-ligand Interactions Based On Chemical Preference & Construction Of Human Cytochrome P450 Substrate Database(CYP-Meta)

In Silico prediction of the new drug-target interactions from existing databases is of important help for the drug discovery processes. Currently, the amount of drug targets that have been identified experimentally is still very small compared with numerous proteins within human genome. In this paper, we have developed a machine learning model based on the chemical-protein interactions from STITCH, a huge database, which is acquired by integrating data from experiments, other databases and text mining protocols. New features from ligand chemical space and interaction networks have been selected as the inputs for support vector machine (SVM) analysis. The 5-fold cross validation and independent analysis show that the high accurancy of prediction ability that outperforms the existing method based on ligand similarity. Moreover, 91 distinct pairs of features have been selected to rebuild a simplifier model, which maintains the same performance as that using all the 322 features. This model is then used to predict potential drug candidates from STITCH database as well as TCM database@Taiwan. The prediction results are then validated either from literatures or experiments. This model is capable of predicting potential new drugs or targets on large scales regardless of protein genomics and structures.The cytochromes P450 (CYP) constitute a superfamily of heme-thiolate enzymes, of which >2700 individual members are currently known are associated with the oxidative metabolism of a large number and variety of organic compounds, both endogenous and exogenous. As far as Homo sapiens are concerned, a total of 57 CYP enzymes have been identified thus far , of which family CYP1, CYP2 and CYP3 are primarily associated with the phase I metabolism of exogenous compounds. A bioinformatics knowledge base (CYP-Meta) collected more than 3000 records of CYPs metabolic information of families CYP1, CYP2, CYP3 and CYP4. Based on these small molecules'impact of the metabolic activity of CYP family members, they are divided into 3 groups: Substrates, Inhibitors and Inducers. So far, the current release of CYP-Meta included about 1500 records of CYP families'metabolic information. In addition to small molecules'basic physical and chemical properties, the database entries also include information on logP, pKa, logD7.4 and relative molecular mass (Mr), together with details of P450-mediated metabolism and the enzymes involved.