CheckM2: a rapid, scalable and accurate tool for assessing microbial genome quality using machine learning.

Area of Science:

• Microbial genomics
• Bioinformatics
• Computational biology

Background:

• Metagenomic sequencing and bioinformatics tools have improved microbial genome recovery.
• Assessing the quality of metagenome-assembled genomes (MAGs) is crucial for downstream analyses.
• Existing quality assessment tools have limitations in accuracy and speed.

Purpose of the Study:

• To introduce CheckM2, an improved machine learning-based method for predicting MAG quality.
• To evaluate CheckM2's performance against existing tools using synthetic and experimental data.
• To demonstrate CheckM2's utility for novel and reduced-genome lineages.

Main Methods:

• CheckM2 utilizes a machine learning approach for genome quality prediction.
• Performance was assessed using both synthetic and real-world metagenomic datasets.
• The CheckM2 database is designed for rapid updates with new reference genomes.

Main Results:

• CheckM2 significantly outperforms existing tools in accuracy and computational speed.
• The tool accurately predicts genome quality for MAGs from novel bacterial and archaeal lineages.
• CheckM2 effectively assesses MAGs from lineages with reduced genome sizes, such as Patescibacteria and DPANN.

Conclusions:

• CheckM2 provides reliable genome quality predictions for MAGs across diverse microbial lineages.
• The tool enhances confidence in biological conclusions drawn from metagenomic data.
• CheckM2 represents a significant advancement in the quality assessment of microbial genomes.