Learning from the unknown: exploring the range of bacterial functionality.

Area of Science:

• Microbial biology
• Bioinformatics
• Computational biology

Background:

• Determining microbial molecular functions is crucial.
• Current methods rely on reference databases and marker genes (e.g., 16S rRNA).
• Limitations exist in current database-dependent approaches.

Purpose of the Study:

• To present a novel, database-independent approach for exploring bacterial functional repertoires.
• To introduce the Fusion scheme for establishing functional relationships and defining Fusion-taxa.
• To develop an alignment-free Siamese Neural Network model for microbial homology detection.

Main Methods:

• Developed a novel Fusion scheme to establish functional relationships between bacteria.
• Utilized functional comparisons, outperforming marker genes for taxonomic assignment.
• Created an alignment-free, nucleic acid-based Siamese Neural Network model using Fusion functions.

Main Results:

• Bacterial functional comparisons using Fusion profiles are superior to marker genes for taxonomic clade assignment.
• Fusion-taxa demonstrate robustness to new organism additions and capture environment-driven diversity.
• The Siamese Neural Network model effectively identifies shared functionality between distant microbial homologs.

Conclusions:

• The Fusion scheme offers a powerful, database-independent method for analyzing bacterial functional repertoires.
• Fusion-taxa provide a robust framework for classifying bacterial diversity, potentially reflecting environmental adaptations.
• This approach enhances the annotation of bacterial functions and deepens the understanding of microbial community structures and interactions.