Nassima Aleb*
Computer Science Department, Jubail University College, Jubail Industrial City, Kingdom of Saudi Arabia
*Corresponding Author: Nassima Aleb, Computer Science Department, Jubail University College, Jubail Industrial City, Kingdom of Saudi Arabia.
Received: November 25, 2021; Published: February 24, 2022
Drug design and discovery is a very challenging and costly process. It involves a crucial phase of drug-target interaction (DTIs) identification. Nevertheless, most existing methods use either binary classification to predict the presence of an interaction in a Drug-Target pair, or regression methods to predict the exact float-value representing the Binding Affinity. These latter methods are more valuable but suffer from unsatisfactory results despite their very sophisticated models and multiple inputs. In this paper, we present a new approach for predicting the strength of drug-target binding, we tackle the question as a Multi-class classification problem. This approach is very rational since the key points, in drug-target interaction, are to have a precise indication about the binding strength and to establish a ranking between drug-target pairs’ binding strengths. Our model input being sequences presenting hidden patterns, we use convolutional LSTM networks, since they inherit the ability in discovering patterns from Convolutional networks, and learning from sequential data from recurrent networks. Besides the usual performance metrics, we investigate new interesting performance metrics that have never been explored before. The results show that our approach is very convincing.
Keywords: Deep Learning Models for Drug Discovery; Drug Repurposing; Drug-target Binding Affinity; Discretization; Multi-Class Classification Models
Citation: Nassima Aleb. “A Deep Learning Multi-class Model for Drug-target Binding Affinity Prediction". Acta Scientific Computer Sciences 4.3 (2022): 14-24.
Copyright: © 2022 Nassima Aleb. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.