Libra: scalable k-mer-based tool for massive all-vs-all metagenome comparisons

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Shotgun metagenomics offers insights into microbial communities but faces limitations with large, complex datasets and database completeness.
• De novo comparative metagenomics allows for direct comparison of metagenomes based on their overall genetic content.

Purpose of the Study:

• To develop a scalable tool for precise, all-vs-all metagenome comparisons.
• To enable large-scale microbial community analyses and link microbial signatures to biological processes.

Main Methods:

• Developed Libra, a tool for all-vs-all metagenome comparison using k-mer content.
• Utilized a scalable Hadoop framework for massive comparisons and Cosine Similarity for distance calculation.
• Implemented Libra in iMicrobe (http://imicrobe.us) via CyVerse cyberinfrastructure.

Main Results:

• Libra performs precise clustering of metagenomes based on k-mer composition and abundance.
• Libra effectively handles massive datasets (80 million to 4.2 billion reads) without data reduction.
• Compared to other methods, Libra maintains high resolution in large-scale comparative analyses.

Conclusions:

• Libra overcomes the resolution limitations of common methods for large metagenomic datasets.
• The tool's scalability and use of all reads enable global-scale microbial analyses.
• Libra facilitates linking microbial community signatures to specific biological functions.