iCYP-MFE: Indentifying Human Cytochrome P450 Inhibitors using Multi-task Learning and Molecular Fingerprint-embedded Encoding
T-H Nguyen-Vo, Q. H. Trinh, L. Nguyen, P-U. Nguyen-Hoang, T-N. Nguyen, D. T. Nguyen, B. P. Nguyen∗ and L. Le∗
The human cytochrome P450 (CYP) superfamily holds responsibilities for the metabolism of both endogenous and exogenous compounds such as drugs, cellular metabolites, and toxins. Inhibition of CYP450 isoforms is closely associated with 1adverse drug reactions which may cause metabolic failures and even induce serious side effects. In modern drug discovery and development, identification of potential CYP isoforms’ inhibitors is highly essential. Besides experimental approaches, numerous computational frameworks have been recently developed to address this biological issue. In our study, we propose robust, stable, and effective prediction models for the virtual screening of five CYP isoforms’ inhibitors, including CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4. The method employs multitask learning combining with molecular fingerprint-embedded encoding to boost the predictive power.
Our results showed that multitask learning had remarkably leveraged useful information from related tasks to promote global performance. In comparison with several state-of-the-art methods, our proposed method is outperforming in all tasks with the highest area under the receiver operating characteristic (ROC) curve (ROC-AUC) of 0.93 and area under the precision-recall (PR) curve (PR-AUC) of 0.92. The evaluated performance once confirms our model’s robustness, stability, and efficiency.
Source code and data are available upon request.