From Machine Learning to Multimodal Models: The AI Revolution in Enzyme Engineering
A research team maps the key tasks that define modern enzyme engineering—function annotation, structure modeling, and property prediction—and explain how AI methods now accelerate each step, from mining candidate enzymes in massive sequence databases to generating new variants with improved activity, stability, or selectivity. From Machine Learning to Multimodal Models: The AI Revolution in Enzyme Engineering Figure 4. Overview of UniKP architecture. (Enzyme sequences encoded into 1024-dimensional embeddings via the ProtT5-XL-UniRef50 protein language model; Substrate sequence transformed into 1024-D feature vectors using a SMILES Transformer; UniKP adopts an Extra Trees regression framework to predict KM,kcat,kcat/KM; Environmental factors (pH/temperature) converted into 1024-D representations. All pathways apply mean pooling to consolidate features while preserving dimensionality. The…