DeepMBEnzy

DeepMBEnzy predicts enzymes capable of mycotoxin biotransformation and compiles these predictions into a searchable database to support mycotoxin detoxification research.

Key Features:

• Enzyme Promiscuity Prediction for Mycotoxin Biotransformation (EPP-MB) model: A fine-tuned pretrained deep learning model that predicts enzyme promiscuity for mycotoxin biotransformation using a cold protein data-splitting approach.
• Large-scale mycotoxin coverage: Predicts potential biotransformation enzymes for over 4000 mycotoxins.
• Database compilation: Archives predicted enzymes and related information for each mycotoxin.
• Validation against experimental data: Achieved 79% validation accuracy against a data set of experimentally confirmed mycotoxin-biotransforming enzymes.

Scientific Applications:

• Mycotoxin detoxification research: Facilitates exploration and utilization of enzyme candidates for mycotoxin biotransformation.
• Intracellular metabolism elucidation: Supports research into understanding the metabolic pathways of mycotoxins within cells.

Methodology:

The EPP-MB model was developed by fine-tuning a pretrained model using a cold protein data-splitting approach, then applied to predict biotransformation enzymes for over 4000 mycotoxins.