DeepAIPs-Pred:\nPredicting Anti-Inflammatory Peptides\nUsing Local Evolutionary Transformation Images and Structural Embedding-Based\nOptimal Descriptors with Self-Normalized BiTCNs

Inflammation is a biological response to harmful stimuli,\nplaying\na crucial role in facilitating tissue repair by eradicating pathogenic\nmicroorganisms. However, when inflammation becomes chronic, it leads\nto numerous serious disorders, particularly in autoimmune diseases.\nAnti-inflammatory peptides (AIPs) have emerged as promising therapeutic\nagents due to their high specificity, potency, and low toxicity. However,\nidentifying AIPs using traditional in vivo methods is time-consuming\nand expensive. Recent advancements in computational-based intelligent\nmodels for peptides have offered a cost-effective alternative for\nidentifying various inflammatory diseases, owing to their selectivity\ntoward targeted cells with low side effects. In this paper, we propose\na novel computational model, namely, <b>DeepAIPs-Pred</b>, for\nthe accurate prediction of AIP sequences. The training samples are\nrepresented using LBP-PSSM- and LBP-SMR-based evolutionary image transformation\nmethods. Additionally, to capture contextual semantic features, we\nemployed attention-based ProtBERT-BFD embedding and QLC for structural\nfeatures. Furthermore, differential evolution (DE)-based weighted\nfeature integration is utilized to produce a multiview feature vector.\nThe SMOTE-Tomek Links are introduced to address the class imbalance\nproblem, and a two-layer feature selection technique is proposed to\nreduce and select the optimal features. Finally, the novel self-normalized\nbidirectional temporal convolutional networks (SnBiTCN) are trained\nusing optimal features, achieving a significant predictive accuracy\nof 94.92% and an AUC of 0.97. The generalization of our proposed model\nis validated using two independent datasets, demonstrating higher\nperformance with the improvement of ∼2 and ∼10% of accuracies\nthan the existing state-of-the-art model using Ind-I and Ind-II, respectively.\nThe efficacy and reliability of <b>DeepAIPs-Pred</b> highlight\nits potential as a valuable and promising tool for drug development\nand research academia.