Coverage

种基因组覆盖度 in the 识别网 workbook = percentage of a species' reference genome covered by raw reads at ≥1× depth (minimap2 + samtools, ported from the original calculate_coverage.py workflow).

Pipeline

top-N species → reference FASTAs (manifest / ref-dir)
reads (raw, pre-QC — same as the legacy workflow)
   → minimap2 -ax <preset> -c … | samtools sort  → BAM
   → samtools depth -a (every position, zero-depth included)
   → per-species: mean depth + % covered at 1x/5x/10x

Depth lines are aggregated in one streaming pass — no numpy arrays, memory stays flat regardless of genome size.

Combined vs per-species

Default (combined): one concatenated reference per sample, one alignment — ~10× faster for multi-species samples. Reads mapping equally well to several species are assigned to one best match only.

--per-species: one alignment per reference — strictly independent estimates, ~N× slower.

Details worth knowing

• Reads are the raw FASTQs (--reads-dir), not the QC/host-removed files — matching the legacy calculate_coverage.py contract, so coverage numbers remain comparable across batches.
• Read length matters: minimap2 -x sr needs ≥50 bp reads; ONT/PacBio data must pass --preset map-ont / --preset map-pb.
• Reference resolution order: manifest by species_taxid → {accession}*.fna in --ref-dir → manifest file_path → {taxid}*.fna glob. Unresolved taxids leave coverage blank.
• --min-depths 1,5,10 controls the thresholds recorded on the 审核明细 sheet; the submission sheet always shows 1×.