A free and open-source implement of the FUGAsseM algorithm (along with installation instructions and demo data) is available from the bioBakery GitHub organization:

github.com/biobakery/fu...

8/8

Specific examples included previously uncharacterized proteins in F. prausnitzii confidently predicted to participate in viral response and CRISPR-Cas systems and Bacteroides thetaiotaomicron proteins newly associated with carbohydrate-metabolizing cassettes:

www.nature.com/articles/s41...
7/8

FUGAsseM accurately predicted uncharacterized proteins with both broadly distributed and species-specific functions across the human microbiome:

www.nature.com/articles/s41...
6/8

We applied FUGAsseM to predict functions for >582k community gene products using 1,595 metagenomes and 800 metatranscriptomes from HMP2. The resulting function catalog is also available as a community resource (see paper for details):

www.nature.com/articles/s41...
5/8

We further quantified the contributional importance of each data modality integrated into the full FUGAsseM model, elucidating their respective impacts on predictive performance:

www.nature.com/articles/s41...
4/8

We comprehensively evaluated FUGAsseM, which exhibited high accuracy in predicting microbial community functional annotations, including those that postdate all training data (temporal holdouts) and those recently validated by experiment:

www.nature.com/articles/s41...
3/8

FUGAsseM integrates community-wide data and demonstrates that metatranscriptomic profiles can accurately predict the functions of previously uncharacterized proteins in microbial communities:

www.nature.com/articles/s41...
2/8