METAMVGL

METAMVGL integrates assembly and paired-end graphs and applies multi-view label propagation to improve metagenomic contig binning by linking fragmented and short contigs and resolving dead-end subgraphs in complex microbial communities.

Key Features:

• Integration of Assembly and Paired-End Graphs: Combines assembly and paired-end graphs to represent contig connectivity and link short or unstable contigs that are difficult to bin using nucleotide composition and read depth alone.
• Automatic Learning of Graph Weights: Automatically learns weights for each graph view to balance their contributions during label inference.
• Multi-View Label Propagation Framework: Employs a uniform multi-view label propagation framework to predict contig labels across integrated graphs.
• Enhanced Connectivity and Error Correction: Increases high-confidence edges in the combined graph to rescue short contigs and correct binning errors associated with dead-end subgraphs.

Scientific Applications:

• Simulated Metagenomic Data: Evaluating binning performance and assembly completeness in controlled simulated metagenomes.
• Mock Communities: Assessing binning accuracy and validation on communities with known composition.
• Real-World Datasets (e.g., Sharon data): Applying to real sequencing datasets such as the Sharon dataset to validate applicability and compare with existing methods.

Methodology:

Combining assembly and paired-end graphs into a unified representation, automatically learning graph weights, and applying multi-view label propagation to propagate labels across views for contig binning.