Predicting functions of uncharacterized gene products from microbial communities.

Area of Science:

• Microbiology
• Bioinformatics
• Genomics

Background:

• The majority of microbial genes lack characterized functions, hindering our understanding of microbial community roles.
• Existing methods for protein function prediction are often limited in scope and scalability for complex microbial communities.

Purpose of the Study:

• To develop and validate a scalable method for inferring putative functions of microbial proteins using community-wide multiomics data.
• To significantly increase the number of characterized protein families within the human gut microbiome.

Main Methods:

• Applied a novel method, FUGAsseM, to analyze multiomics data from gut metagenomes and metatranscriptomes.
• Integrated sequence similarity, genomic proximity, and domain-domain interactions to improve functional prediction accuracy.
• Predicted functions for over 443,000 protein families, including those with weak or no homology to known proteins.

Main Results:

• Successfully predicted high-confidence functions for 82.3% of analyzed protein families (>443,000 total).
• Identified functions for thousands of previously uncharacterized protein families, including those lacking homology.
• Demonstrated FUGAsseM's generalizability across diverse microbial communities and its comparable accuracy to single-organism methods.

Conclusions:

• The FUGAsseM method significantly expands the functional annotation of microbial proteins, particularly in undercharacterized communities.
• This work provides a powerful tool for exploring the functional landscape of microbial ecosystems, starting with the human gut microbiome.