DLMM-DTI: Deep Learning-Based Mamba Model for Drug-Target Interaction Binary Classification Prediction

Drug–target interactions (DTIs) play a vital role in drug development, as accurately predicting them can speed up discovery and lower costs. However, traditional DTI identification methods are costly and time-intensive, mainly due to the complex nature of molecular and protein spaces. Furthermore, existing deep learning approaches overlook necessary molecular substructure, leading to suboptimal predictive accuracy. To address these challenges, we propose DLMM-DTI, a novel prediction model that integrates BCM, CFM, CNN, and Mamba techniques to capture diverse features of drug and protein fragments. By incorporating multiple representations, DLMM-DTI effectively captures both intra- and inter-model interactions. Experimental results demonstrate that DLMM-DTI outperforms existing methods on Human, Celegans and BIOSNAP datasets, reducing training costs and improving prediction accuracy.