Deep Learning-Driven\nInsights into Enzyme–Substrate\nInteraction Discovery

Enzymes are ubiquitous catalysts with enormous application\npotential\nin biomedicine, green chemistry, and biotechnology. However, accurately\npredicting whether a molecule serves as a substrate for a specific\nenzyme, especially for novel entities, remains a significant challenge.\nCompared with traditional experimental methods, computational approaches\nare much more resource-efficient and time-saving, but they often compromise\non accuracy. To address this, we introduce the molecule–enzyme\ninteraction (MEI) model, a novel machine learning framework designed\nto predict the probability that a given molecule is a substrate for\na specified enzyme with high accuracy. Utilizing a comprehensive data\nset that encapsulates extensive information on enzymatic reactions\nand enzyme sequences, the MEI model seamlessly combines atomic environmental\ndata with amino acid sequence features through an advanced attention\nmechanism within a hierarchical neural network. Empirical evaluations\nhave confirmed that the MEI model outperforms the current state-of-the-art\nmodel by at least 6.7% in prediction accuracy and 8.5% in AUROC, underscoring\nits enhanced predictive capabilities. Additionally, the MEI model\ndemonstrates remarkable generalization across data sets of varying\nqualities and sizes. This adaptability is further evidenced by its\nsuccessful application in diverse areas, such as predicting interactions\nwithin the CYP450 enzyme family and achieving an outstanding accuracy\nof 90.5% in predicting the enzymatic breakdown of complex plastics\nwithin environmental applications. These examples illustrate the model’s\nability to effectively transfer knowledge from coarsely annotated\nenzyme databases to smaller, high-precision data sets, robustly modeling\nboth sparse and high-quality databases. We believe that this versatility\nfirmly establishes the MEI model as a foundational tool in enzyme\nresearch with immense potential to extend beyond its original scope.