Explainable Artificial Intelligence to Unveil Patterns of Antioxidant Peptides for Free Radical Regulation

Abstract Antioxidant peptides are emerging as promising therapeutic and nutraceutical agents because of their capacity to neutralize free radicals, modulate oxidative stress, and contribute to disease prevention in fields ranging from oncology to dermatology and food science. However, traditional computational approaches often fail to capture the biochemical semantics and evolutionary context of peptide sequences, limiting the predictive accuracy and mechanistic interpretability. Here, we present an explainable deep learning framework integrating evolutionary scale modeling (ESM) embeddings with a temporal convolutional network (TCN) and long short-term memory (LSTM) architecture to classify antioxidant peptides and uncover key sequence determinants of activity. ESM embeddings provide rich evolutionary and structural priors, enabling the model to capture both local k-mer motifs and global sequence dependencies. To ensure transparency, we employed a multimethod interpretability strategy using Anchor, LIME, and SHAP, which identified robust, potentially biologically meaningful hypotheses containing residues rich in cysteine (C), arginine (R), lysine (K), histidine (H), aspartic acid (D), glutamic acid (E), threonine (T), glutamine (Q), proline (P), methionine (M), tyrosine (Y), and tryptophan (W) as central to antioxidant activity. Anchor offered deterministic, high-precision motif rules, while LIME and SHAP highlighted redundant and context-dependent contributions, together forming a layered interpretability framework. This study not only achieves state-of-the-art classification performance but also bridges black-box modeling and biochemical understanding, offering a generalizable, hypothesis-generating platform for rational peptide design and discovery that is important in human health.