plASgraph2: using graph neural networks to detect plasmid contigs from an assembly graph

• DOI: 10.3389/fmicb.2023.1267695
• Publication Date: 2023-10-06T09:51:11Z
• Authors: Janik Sielemann; Katharina Sielemann; Broňa Brejová; Tomáš Vinař; Cedric Chauve
• Keywords: plasmids; classification; machine learning (ML); bioinformatics; Microbiology; assembly graph; QR1-502; 3. Good health

Identification of plasmids from sequencing data is an important and challenging problem related to antimicrobial resistance spread other One-Health issues. We provide a new architecture for identifying plasmid contigs in fragmented genome assemblies built short-read data. employ graph neural networks (GNNs) the assembly propagate information nearby nodes, which leads more accurate classification, especially short that are difficult classify based on sequence features or database searches alone. trained plASgraph2 set samples ESKAPEE group pathogens. either outperforms performs par with wide range state-of-the-art methods testing sets independent On one hand, our study provides easy use tool contig classification bacterial isolates; it serves as proof-of-concept GNNs genomics. Our software available at https://github.com/cchauve/plasgraph2 training https://github.com/fmfi-compbio/plasgraph2-datasets.