Drug Target Interaction Prediction using Graph Convolution based Neural Fingerprinting
Adarsh Joshy, Govindu CJVS Kasyap, Puchakayala Dheeraj Reddy, I. T. Anjusha, K. A. Abdul Nazeer · 2022 IEEE 19th India Council International Conference (INDICON) · 2022
The time and cost invested for development of new drugs for orphan diseases is substantial and it is not profitable as it only serves a small patient population. This process can be simplified using Drug Repurposing, a method where clinically approved drugs are predicted to have an impact in the treatment of novel diseases. A single drug may treat multiple diseases by acting on its related corresponding targets. This idea can be leveraged to repurpose drugs for better treatment of many diseases. Network models and text mining are some of the general approaches used to find repurposable drugs, but being rule based, they fail in ambiguous cases. In order to overcome these shortcomings, we propose a Deep Learning based approach to precisely predict drugs with maximum Drug Target Interaction (DTI) using KIBA dataset, which contains known drug-target interaction pairs. Having shown notable successes in Cancer treatment, drug repurposing can be extrapolated for many other conditions. In this paper, we predict drug target interaction using Neural Fingerprinting. This approach efficiently embeds the drug molecular structures and thereby predicts the binding affinity (Kd) with a MSE of 0.197 on the KIBA dataset.