In silico modelling, identification of crucial molecular fingerprints, and prediction of new possible substrates of human organic cationic transporters 1 and 2†
Abstract
There is a great demand to utilize different in silico tools to address unwanted drug metabolism and pharmacokinetics issues in drug discovery. There is also a demand to optimize existing drug therapeutics by understanding their interactions with various transporters in the body. The cation membrane transporter is among one of the major crucial transporters within the body to regulate movement of foreign molecules/drugs across the cell membrane. The prime objective of this study is to find out the structural fingerprints within molecules to be recognized as substrate/non-substrate against human organic cation transporters (hOCTs). This study may pave the way for a more detailed understanding of the physiological and pharmacological roles of transporters as well as in predicting pharmacokinetics and pharmacodynamics in the design and development of better cationic drugs. The in silico study involving physicochemical parameter analysis revealed a trend that was distinct for substrate and non-substrate molecules present in the dataset. A hyperfine geometric distribution based strategy was also utilized for obtaining the detailed distribution of different functional moiety in substrates and non-substrates for hOCTs. Classification QSAR study by Monte-Carlo optimization and Bayesian modelling methods were used for the identification of crucial structural fingerprints important for substrate activity. Classification QSAR study with the help of Monte-Carlo optimization in case of hOCT1 revealed important structural attributes like aromatic rings with branching, two aromatic rings at one bond distance, presence of sulphur atom in the compound, etc. These seemed to have vital roles of hOCT1 substrate activity. Similarly, a Bayesian classification model was also generated which gave important fingerprints for the analysis of substrates in the case of hOCT1. Models built for hOCT2 by Monte Carlo optimization and Bayesian classification study have not given significant results from the dataset. Finally, various machine learning methods (kNN, ASNN, WEKA RF, XG-BOOST) with the combination of various descriptors were utilized for the development of models and consensus models were generated separately using the best models for both hOCT1 and hOCT2. The consensus models were utilized to predict some recently FDA approved drugs for possibly hOCT1 and hOCT2 substrates. To confirm further, the predicted compounds were docked and the analysis showed that they may bind to the different binding sites of the transporter.