Unlocking Enzyme Discovery: A High-Resolution Gene Cluster Database Powered by Phylogenetic Insights and Machine...

Area of Science:

• Biotechnology
• Bioinformatics
• Enzyme Engineering

Background:

• Vast genomic data requires better annotation for enzyme discovery.
• Current methods struggle to identify novel, high-value enzymes efficiently.

Purpose of the Study:

• To develop and validate an integrated computational framework for large-scale enzyme discovery.
• To enhance the identification and prioritization of functional enzyme variants.

Main Methods:

• Constructed a cross-kingdom phylogenetic database.
• Employed multilocus phylogeny for candidate mining.
• Utilized a protein language model for activity prediction.
• Applied multilevel residue-atom contact rescoring for false positive removal.

Main Results:

• Discovered numerous undocumented enzyme homologues in the r-BOX pathway.
• Achieved R² = 0.68 for activity prediction, outperforming UniKP.
• Improved early enrichment (EF1%) by 16-fold using contact scoring.
• Increased FadB enzyme titers from 0.65 g/L to 10.2 g/L through experimental validation.

Conclusions:

• The developed framework is robust and generalizable for discovering scarce, high-value enzymes.
• The pipeline effectively prioritizes functional enzyme variants at scale.
• Demonstrated significant improvements in enzyme production via experimental validation.