Learn an error model from pre-computed sample JSON files.
dada2-rs sample reads/*.fastq.gz -o sampled/ --gzip
dada2-rs errors-from-sample sampled/*.json.gz --threads 24 -o err.jsonReads JSON files produced by sample (or derep) and
iteratively runs the DADA2 algorithm, re-fitting the chosen error model until
self-consistency. Output is the same three flat 16 × nq matrices as
learn-errors: trans, err_in, err_out.
This is the two-step form of learn-errors. The difference matters: sample
controls what is accumulated, so pre-sampling each file and then fitting here
spreads learning across samples rather than letting one deep sample fill the
whole budget. If you are reaching for learn-errors --nbases and worrying about
which samples it actually saw, this pair is the answer.
<INPUT>... — sample or derep JSON files (.json / .json.gz).
There is no --nbases / --randomize / --phred-offset here; those belong to
the sample step that produced the input.
Identical in meaning to learn-errors. In particular
--omega-c defaults to 0 here too, matching R DADA2's learnErrors().
--threads (default 1) — threads for parallel sample processing.
--output / -o — write JSON here instead of stdout.
--compact — minified JSON.
learn-errors— the one-step formsample- learn-errors --nbases convergence