Troubleshooting common errors in assemblies of long-read metagenomes.

Area of Science:

• Genomics
• Bioinformatics
• Metagenomics

Background:

• Assessing long-read assembly accuracy in complex environmental metagenomes is difficult.
• Underrepresented organisms pose additional challenges for genome recovery.

Purpose of the Study:

• To benchmark four leading long-read assembly software programs.
• To identify and quantify errors in metagenome assemblies.
• To develop a reproducible workflow for assembly evaluation.

Main Methods:

• Benchmarking HiCanu, hifiasm-meta, metaFlye, and metaMDBG on 21 PacBio HiFi metagenomes.
• Quantifying read clipping events during mapping to assembled contigs.
• Analyzing mock communities, gut microbiomes, and ocean samples.

Main Results:

• Long-read metagenome assemblies can contain over 40 errors per 100 million base pairs.
• Identified errors include multi-domain chimeras, premature circularization, haplotyping errors, excessive repeats, and phantom sequences.
• Significant divergence between assembled contigs and source reads was observed.

Conclusions:

• Current long-read assembly methods exhibit substantial error rates in metagenomic contexts.
• A novel open-source tool and workflow facilitate rigorous evaluation of assembly errors.
• Improved genome recovery from metagenomes requires addressing these identified assembly inaccuracies.