Predicting the Activity of Mutation-Specific CYP450

Abstract

Pharmacogenomics is a field of research that studies how human genes respond to drugs. If a patient suffers from the same disease during drug treatment, the same drugs are prescribed, but the effects of the drug response are known to vary from the patient. The enzyme most involved in drug metabolism is cytochrome P450(CYP450). One of the major characteristics of CYP450 is that it has a variety of genetic variations. Therefore, it is important to understand the activity of the mutations related to drug responses for effective treatment. However, it is time-consuming to identify the drug responses considering all drugs and alleles via biochemical experiments. Thus, the aim of this paper is to predict the activity of the CYP450 enzyme for specific drugs using deep learning approaches to achieve better prediction accuracy. The result of proposed model showed that better performance than other baseline models and have strength in unseen alleles (external dataset) considering PCC, 𝑅2 and MAE. In particular, it is expected to help predicting patient’s drug activity using the proposed model, given that validation of CYP2B6, novel variants, showed good prediction results