PLASMe: a tool to identify PLASMid contigs from short-read assemblies using transformer.

Area of Science:

• Genomics
• Bioinformatics
• Microbiology

Background:

• Plasmids are crucial mobile genetic elements facilitating bacterial horizontal gene transfer.
• Next-generation sequencing (NGS) is key for plasmid discovery, but assembly artifacts like contigs complicate detection.
• Existing plasmid detection tools have limitations, missing divergent plasmids or having low precision.

Purpose of the Study:

• To develop a robust plasmid detection tool, PLASMe, that overcomes limitations of current methods.
• To enhance the accuracy and reliability of plasmid contig identification from diverse sequencing data.

Main Methods:

• PLASMe integrates alignment-based methods for closely related plasmids and order-specific Transformer models for divergent plasmids.
• Plasmid sequences are encoded as a language using protein cluster-based tokens for Transformer analysis.
• Performance was evaluated on simulated (CAMI2) and real metagenomic/plasmidome datasets.

Main Results:

• PLASMe achieved the highest F1-score in detecting complete plasmids and plasmid contigs compared to other tools.
• The tool demonstrated reliable performance on real-world metagenomic and plasmidome data.
• Analysis of marker genes confirmed PLASMe's superior reliability.

Conclusions:

• PLASMe offers a significant advancement in plasmid detection accuracy and reliability.
• The tool's hybrid approach effectively addresses challenges posed by fragmented NGS data.
• PLASMe is a valuable asset for cataloging plasmids and understanding their roles in bacterial evolution and gene transfer.