TNFIPs-Net: A deep learning model based on multi-feature fusion for prediction of TNF-α inducing epitopes

Tumor necrosis factor alpha (TNF-α) is a cytokine belonging to the tumor necrosis factor family. It plays a crucial regulatory role in the immune system and is involved in various biological processes. TNF-α inducing epitopes are specific antigenic epitopes capable of stimulating or inducing the production of TNF-α in cells. By identifying and studying TNF-α inducing epitopes, we gain a better understanding of their relevance to diseases, offering novel targets and intervention strategies for drug development and treatment. Furthermore, in-depth research on TNF-α inducing epitopes provides important insights and guidance for personalized medicine and precision immune therapy. In this study, we propose a novel deep learning model based on multi-feature fusion for predicting TNF-α inducing epitopes (TNFIPs-Net). Our model utilizes a dual-branch architecture guided by adaptive features and hand-crafted features. In the encoder layers, the adaptive word embedding features and hand-crafted features are separately input to different encoders (transformer and BiGRU, respectively). Finally, an attention mechanism is employed in the output layer to fuse the features from both branches. Through experimental validation, the fusion of multiple features and the use of self-attention mechanism enable the model to capture complex feature information more effectively, thereby improving the predictive performance of TNF-α inducing epitopes. The datasets and code used in this research are available at https://github.com/yujiexu321/TNFIPs-Net.