Accurate plasmid reconstruction from metagenomics data using assembly-alignment graphs and contrastive learning

Area of Science:

• Genomics
• Microbiology
• Bioinformatics

Background:

• Plasmids are vital extrachromosomal DNA elements facilitating bacterial horizontal gene transfer, often conferring antibiotic resistance.
• Plasmids are significantly underrepresented in genomic databases due to assembly challenges like mosaicism and microdiversity.
• Existing plasmid assemblers struggle with fragmented, entangled, and low-coverage assembly graphs from single samples.

Purpose of the Study:

• To develop a novel computational method for improved plasmid and cellular genome recovery from metagenomic samples.
• To address the limitations of current plasmid assembly tools by leveraging cross-sample data integration.
• To enhance the study of organism-plasmid associations and intraplasmid diversity.

Main Methods:

• Introduction of PlasMAAG (plasmid and organism metagenomic binning using assembly-alignment graphs).
• Generation of an 'assembly-alignment graph' by complementing single-sample assembly graphs with cross-sample signals.
• Integration of the assembly-alignment graph with standard binning features for enhanced plasmid reconstruction.

Main Results:

• PlasMAAG reconstructed 50-121% more near-complete plasmids on synthetic datasets compared to existing methods.
• Achieved a 28-106% improvement in the Matthews correlation coefficient for geNomad contig classification.
• Reconstructed 33% more plasmid sequences from hospital sewage samples, outperforming competing methods.

Conclusions:

• PlasMAAG significantly advances plasmid recovery and characterization from complex metagenomic data.
• The method improves the accuracy of plasmid identification and classification.
• Enables deeper insights into plasmid dynamics, host associations, and diversity within microbial communities.