CoCoPyE: feature engineering for learning and prediction of genome quality indices

Area of Science:

• Bioinformatics
• Genomics
• Microbial Ecology

Background:

• Metagenomic sequencing enables microbial genome reconstruction, but data quality varies widely.
• Accurate quality assessment of these genomes is crucial for downstream analyses.
• Current methods often rely on limited single-copy genes, impacting accuracy.

Purpose of the Study:

• To develop a novel, accurate tool for assessing the quality of metagenome-assembled genomes.
• To improve upon existing methods for estimating genome completeness and contamination.

Main Methods:

• CoCoPyE utilizes a two-stage approach: genomic marker identification followed by machine learning refinement.
• The tool employs a novel feature extraction and transformation scheme.
• A freely available Python implementation facilitates integration into existing pipelines.

Main Results:

• CoCoPyE demonstrated more accurate prediction of quality indices in simulation studies compared to existing tools.
• The tool provides a fast and efficient method for genome quality assessment.
• A web server is available for easy access and testing.

Conclusions:

• CoCoPyE offers a new, improved method for evaluating metagenome-assembled genome quality.
• It complements and enhances existing genome quality assessment tools.
• Researchers can use CoCoPyE to confidently differentiate between high-quality and low-quality genome assemblies.