pku-embl/metagenomics-skills/coverm

CoverM

Upstream: wwood/CoverM

Citation

Aroney, S. T. N. et al. CoverM: read alignment statistics for metagenomics. Bioinformatics 41, btaf147 (2025). https://doi.org/10.1093/bioinformatics/btaf147

See also docs/references.md.

Analytical thinking

CoverM unifies per-contig (binning inputs) and per-genome (community profiling) coverage using Mosdepth-style arrays. Relative abundance (default genome mode) scales mean coverages by the fraction of reads that align — incomplete references leave many reads unmapped; CoverM accounts for that in its relative-abundance formula.

Important defaults / assumptions from the paper:

• Genome presence filter: by default ≥10% of genome length covered ≥1× before

assigning non-zero coverage (limits off-target hits in conserved regions).

• Relative abundance assumes roughly similar average genome sizes for mapped vs

missing community members, and roughly complete/uncontaminated references.

• Near-identical references distort recruitment — CoverM can dereplicate with

Galah using skani/FastANI/Dashing ANI before mapping.

• Aligners: strobealign / minimap2 / BWA; filter by identity, alignment length,

or fraction of read aligned.

Relative abundance ≠ absolute load. Use the same dehosted reads that define biological samples. Merge per-sample outputs carefully (headers differ by version).

How to run

# genome relative abundance vs dereplicated MAGs
coverm genome --coupled ${id}_1.fastq ${id}_2.fastq \
  --genome-fasta-directory drep95/dereplicated_genomes/ -x fa \
  -t 8 -o temp/coverm/${id}.txt
 
# contig coverage for binning (example)
coverm contig -r contigs.fa --coupled ${id}_1.fastq ${id}_2.fastq \
  -t 8 -o temp/coverm/${id}.contig.tsv

Key parameters

Validation, recovery, and provenance

Require a non-empty table with unique sample/reference labels, finite non-negative values, and all intended samples. For contig binning, verify contig IDs exactly match the assembly and confirm the selected method/column layout is accepted by the binner. For genome abundance, report covered fraction, mapped/unmapped fraction, presence filter, identity/aligned-fraction filters, and multi-mapping policy; inspect unexpected zero/high values.

Rerun into a new output after changing catalogue, mapper, or filters. Capture coverm --version, complete command, read and reference checksums, dereplication manifest, mapper/version, table schema, and mapping logs. A non-zero genome abundance is a recruitment estimate, not proof of organism viability or absolute load.

Related skills

drep · skani · strobealign · gtdbtk · visualization · metabat2