Predicting oxygen levels in microbial habitats using a metagenome-based approach

Area of Science:

• Microbial Ecology and Genomics
• Bioinformatics and Computational Biology

Background:

• Oxygen is a critical factor influencing microbial life, but direct in situ measurement is challenging.
• Bacteria exhibit diverse oxygen preferences, making them potential bioindicators of oxygen levels.
• Quantifying bacterial community oxygen preferences can infer environmental oxygen variations and metabolic strategies.

Purpose of the Study:

• To develop a computational tool, OxyMetaG, for predicting oxygen availability using metagenomic data.
• To leverage bacterial gene content as a proxy for oxygen levels in various environments.
• To provide a method for inferring past and present oxygen conditions where direct measurement is not feasible.

Main Methods:

• Ensemble machine learning identified 20 key genes predicting bacterial oxygen tolerance.
• A relationship was established between aerobic:anaerobic gene abundance ratios and aerobic bacteria proportions.
• OxyMetaG was developed to process metagenomic reads, map to indicator genes, and predict oxygen levels (0-100%).

Main Results:

• OxyMetaG successfully predicted oxygen levels in diverse environmental and host-associated metagenomes.
• Application to 540 surface soils revealed predominantly oxic conditions, with reduced oxygen in wetter, finer-textured sites.
• Analysis of 73 human gut samples showed a significant decrease in oxygen levels from infancy to early childhood.

Conclusions:

• OxyMetaG offers a robust method for estimating oxygen availability directly from shotgun metagenomic data.
• The tool is effective even with low sequencing depth and avoids computationally intensive genome assembly.
• OxyMetaG has broad applicability for characterizing oxygen dynamics in microbial ecosystems across different scales and time periods.