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

BMC Bioinformatics. 2019 Oct 3;20(1):486. doi: 10.1186/s12859-019-3068-y.


Background: Recent advances in high-volume sequencing technology and mining of genomes from metagenomic samples call for rapid and reliable genome quality evaluation. The current release of the PATRIC database contains over 220,000 genomes, and current metagenomic technology supports assemblies of many draft-quality genomes from a single sample, most of which will be novel.

Description: We have added two quality assessment tools to the PATRIC annotation pipeline. EvalCon uses supervised machine learning to calculate an annotation consistency score. EvalG implements a variant of the CheckM algorithm to estimate contamination and completeness of an annotated genome.We report on the performance of these tools and the potential utility of the consistency score. Additionally, we provide contamination, completeness, and consistency measures for all genomes in PATRIC and in a recent set of metagenomic assemblies.

Conclusion: EvalG and EvalCon facilitate the rapid quality control and exploration of PATRIC-annotated draft genomes.

Keywords: CheckM; Genome annotation; Genome quality; Machine learning; Metagenomics; RAST; Random forest; Supervised learning.

MeSH terms

  • Databases, Genetic*
  • Genome, Archaeal*
  • Genome, Bacterial*
  • Machine Learning*
  • Metagenomics / methods*
  • Metagenomics / standards
  • Software