Quantification modes

oarfish supports five ways of providing input for bulk quantification. All of them feed the same EM-based quantification core; they differ only in how alignments to the transcriptome are obtained. The mode is selected implicitly by which input flag you pass. (Single-cell quantification, --single-cell, is separate and accepts only a transcriptome BAM — see Single-cell mode.)

Mode	Selected by	What it does
Alignment (transcriptome BAM)	--alignments	Quantify pre-computed alignments of reads to the transcriptome (a name-sorted BAM, e.g. from minimap2/pbmm2).
Read (transcriptome mapping)	--reads + --annotated/--index	Map raw reads to the transcriptome internally (with the built-in rammap mapper, or spearfish with --mapper spearfish), then quantify.
RAD (spearfish mappings)	--alignments with a RAD file	Quantify the mappings in a RAD file written by spearfish (spearfish map --rad). The file is recognized by its contents, and carries the transcripts, the score scale, the preset and whether spans reach the read ends, so no other options are needed.
Genome read-projection	--reads + --genome + --annotation	Spliced-align raw reads to the genome, project each read’s genomic alignments onto the annotated transcripts (via bramble), then quantify.
Genome-alignment projection	--genome-alignments + --annotation	Project an existing spliced genome BAM (e.g. from minimap2 -ax splice) onto the annotated transcripts, then quantify.

Which should I use?

The transcriptome read/alignment modes are the most mature and are the right default when you have (or can build) a transcriptome reference. The genome-based modes are useful when you want to align against the genome — for example to keep one genomic alignment pass for multiple downstream analyses, to start from genome BAMs you already have, or to capture reads that fall outside annotated transcript boundaries (recovered via soft-clip rescue). On our benchmarks, genome read-projection quantification closely matches direct transcriptome quantification when rescue is enabled (see expectations).

For details of each input, see Input; for worked commands, see Usage examples.