METAMVGL: a multi-view graph-based metagenomic contig binning algorithm by integrating assembly and paired-end graphs

Area of Science:

• Microbiology
• Bioinformatics
• Genomics

Background:

• De novo assembly of next-generation sequencing data often yields incomplete microbial genomes due to community complexity.
• Metagenome assembly binning groups fragmented contigs into potential genomes using nucleotide composition and read depth.
• Current binning methods struggle with short contigs, as their features are less stable.

Purpose of the Study:

• To develop a novel metagenomic contig binning algorithm that improves the handling of short contigs and corrects binning errors.
• To integrate both assembly graphs and paired-end (PE) graphs for more robust genome reconstruction.

Main Methods:

• Developed METAMVGL, a multi-view graph-based algorithm utilizing label propagation.
• Integrated assembly graphs and PE graphs, learning graph weights automatically.
• Employed a uniform multi-view label propagation framework for contig classification.

Main Results:

• METAMVGL effectively rescued short contigs and corrected binning errors, particularly at dead ends of the assembly graph.
• The algorithm leveraged more high-confidence edges from the combined graph structure.
• METAMVGL demonstrated superior performance compared to state-of-the-art tools like MaxBin2, MetaBAT2, and CONCOCT on simulated, mock, and real infant fecal metagenomic data.

Conclusions:

• METAMVGL significantly enhances the binning of short contigs in metagenomic datasets.
• The developed tool outperforms existing contig binning algorithms, offering improved accuracy and completeness for microbial genome assembly.