Metagenomic screening overview

Environmental / contamination profile

A 2,000,000-read subsample of the reads that didn't align to the C. arabica reference, classified against Kraken2's PlusPF-8 database (bacteria, archaea, viruses, protozoa, fungi, human).

Bars below show the species/genus-level taxa Kraken2 could actually pin down, within that classified slice — the higher ranks above them in the tree (Bacteria, Pseudomonadota, ...) are cumulative parents of these and are left out here to avoid double-counting.

Direct coffee-match check

The same subsample, classified instead against a dedicated database built from just three genomes: C. arabica (the target) and its two diploid parent species C. canephora and C. eugenioides (arabica is a natural hybrid of the two). This asks directly: does any of the off-target read pool actually look like coffee?

Are those coffee-classified reads real?

A k-mer hit isn't the same as a confirmed match — Kraken2 can classify a read from a single 35bp k-mer match. Every Coffea-classified hit was re-aligned to the C. arabica reference with the pipeline's own bwa parameters to check directly, then re-aligned again with a soft-clip-tolerant aligner (bwa mem) and clustered by genomic position to see exactly where on the reference they're landing.