Quantification

spearfish reports scored transcript sets; an EM turns them into abundances. oarfish does this, with coverage and positional models. It needs a build with spearfish support: its --help lists spearfish as a --mapper.

In one step: oarfish maps with spearfish

oarfish --reads reads.fastq.gz --annotated transcripts.fa.gz --seq-tech ont-cdna \
    --mapper spearfish -j 16 -o sample

oarfish builds a spearfish index from the transcripts in process (or opens one written by spearfish index, given with --index transcripts.sfi), maps the reads with the preset that --seq-tech names (ont-cdna, ont-drna or pac-bio-hifi), and quantifies them, writing sample.quant and sample.meta_info.json. It sets the mapping options its models need itself: with --model-truncation, for instance, read spans are extended to the read ends. This is the recommended route.

In two steps: map to RAD, then quantify

spearfish map transcripts.sfi reads.fastq.gz --preset ont -t 16 --rad sample.rad
oarfish -a sample.rad -o sample -j 16

The RAD file holds every read’s scored hits, and oarfish takes the transcripts, the score scale, the preset and whether spans reach the read ends from it, so no other options are needed. Map this way to keep the mappings: to quantify them several ways without mapping again, or to map and quantify on different machines. Mapping settings are yours to choose here, so map with --extend-ends for oarfish’s truncation model (--model-truncation); the ont preset already extends them.