A machine learning-based service for estimating quality of genomes using PATRIC

Area of Science:

• Microbiology
• Bioinformatics
• Genomics

Background:

• High-volume sequencing and metagenomic data necessitate efficient genome quality assessment.
• The PATRIC database houses over 220,000 genomes, with ongoing generation of novel draft genomes from metagenomic assemblies.

Purpose of the Study:

• To introduce and evaluate two new tools, EvalG and EvalCon, for rapid genome quality control.
• To assess the performance and utility of these tools within the PATRIC annotation pipeline.

Main Methods:

• EvalCon employs supervised machine learning to compute an annotation consistency score.
• EvalG utilizes a modified CheckM algorithm to estimate genome contamination and completeness.
• These tools were applied to all genomes in PATRIC and a set of metagenomic assemblies.

Main Results:

• Performance metrics for EvalG and EvalCon were reported.
• Contamination, completeness, and consistency measures were generated for extensive genome datasets.
• The utility of the annotation consistency score was evaluated.

Conclusions:

• EvalG and EvalCon significantly enhance the quality control process for draft genomes.
• These tools facilitate rapid exploration and assessment of large-scale genomic data in PATRIC.