Discovery of Toxin-Degrading Enzymes with Positive Unlabeled Deep Learning

Abstract

Identifying functional enzymes for the catalysis of specific biochemical reactions is a major bottleneck in the de novo design of biosynthesis and biodegradation pathways. Conventional methods based on microbial screening and functional metagenomics require long verification periods and incur high experimental costs; recent data-driven methods apply only to a few common substrates. To enable rapid and high-throughput identification of enzymes for complex and less-studied substrates, we propose a robust enzyme's substrate promiscuity prediction model based on positive unlabeled learning. Using this model, we identified 15 new degrading enzymes specific for the mycotoxins ochratoxin A and zearalenone, of which six could degrade >90% mycotoxin content within 3 h. We anticipate that this model will serve as a useful tool for identifying new functional enzymes and understanding the nature of biocatalysis, thereby advancing the fields of synthetic biology, metabolic engineering, and pollutant biodegradation.

Enhanced Deep-Learning Model for Carbon Footprints of Chemicals

“…However, it could not provide clear insights on, for example, PCF-intensive substructures and raw materials to guide the design of more sustainable molecules and processes. The attention mechanism has been used to identify important features that affect the activity of molecules, enzymes, and reactions. , FineChem 2 employs the attention mechanism to capture the contributions of substructures in a chemical to its PCF, thereby promoting learned knowledge and improving the model interpretability.…”

“…The attention mechanism has been used to identify important features that affect the activity of molecules, enzymes, and reactions. 22 , 49 FineChem 2 employs the attention mechanism to capture the contributions of substructures in a chemical to its PCF, thereby promoting learned knowledge and improving the model interpretability.…”

“…The contamination of proxy data is transmitted to these pre-LCA tools, compromising their performance. Furthermore, while ML has been successfully applied in many fields, interpretability remains challenging for ML-based models. , Previous tools used Shapley additive explanations (SHAP) to identify the critical molecular descriptors that are relevant to PCFs. However, the detailed correlations between PCFs and functional groups and substructures cannot be distinctly quantified, which limits their application in the design of more sustainable molecules.…”

Comprehensive Review of Aflatoxin and Ochratoxin A Dynamics: Emergence, Toxicological Impact, and Advanced Control Strategies

“…They elaborate on leveraging computational biology for searching genomic databases to identify candidate organisms with potential mycotoxin detoxification capabilities, which can be further enhanced via synthetic biology techniques to optimize the expression and activity of these enzymes. For instance, Zhang et al [ 241 ] utilized a positive unlabeled deep learning approach to identify new enzymes capable of degrading mycotoxins such as OTA, demonstrating the potential of machine learning to predict enzyme-substrate specificity in complex substrates traditionally challenging for experimental approaches. This method, leveraging data from extensive biochemical databases, underscores a shift towards rapid, high-throughput screening methods that could revolutionize the discovery of biodegrading enzymes in synthetic biology frameworks.…”