Fingerprinting cities: differentiating subway microbiome functionality.

Area of Science:

• Microbiology
• Bioinformatics
• Environmental Science

Background:

• Human microbiome composition is increasingly linked to public and individual health.
• City subway systems serve as high-density environments facilitating microbial exchange among passengers.
• Previous research on subway microbiomes primarily focused on taxonomic composition, lacking explicit functional analysis.

Purpose of the Study:

• To functionally profile subway microbiomes from diverse cities.
• To develop a predictive model for determining a city's origin based on its subway microbiome.
• To assess the impact of data balancing on model performance and identify city-specific functional signatures.

Main Methods:

• Utilized the mi-faser tool for high-speed functional annotation of shotgun metagenomic data.
• Applied machine learning to build a predictive model for city origin from microbiome functional profiles.
• Implemented data balancing techniques to mitigate biases in training and testing datasets.

Main Results:

• Achieved high accuracy (76-94%) in predicting city of origin using a balanced functional profiling model.
• Demonstrated that unbalanced data leads to over-estimated performance (90-97%).
• Identified distinct functional signatures, such as mycobacterial depletion in New Zealand and photosynthesis-related functions in New York, Porto, and Tokyo.

Conclusions:

• The mi-faser method enables rapid and effective functional analysis of large metagenomic datasets.
• Balanced data is crucial for reliable microbiome-based predictions and avoids performance overestimation.
• Functional analysis of subway microbiomes reveals both shared characteristics and unique microbial signatures across geographically diverse cities.