Built for long reads
Presets for ONT cDNA, ONT direct RNA and PacBio HiFi. Seeds from locally consistent parsing survive the errors of long reads, and read ends are clipped where adapters hang off them.
spearfish
Fast long-read transcriptome mapping. Every isoform a read is compatible with, scored, with each piece of evidence computed once instead of once per isoform.
What spearfish does
Section titled “What spearfish does”The isoforms of a gene are built from a shared pool of exons, so a long read is usually compatible with many transcripts at once. Mappers designed for genomes treat each transcript as a separate target and repeat the same seeding, chaining and alignment for every isoform that contains an exon. spearfish does that work once per distinct piece of sequence, and reports, for each read, every compatible transcript with its position and its score relative to the best. We call this delta mapping.
Delta mapping
Each candidate transcript gets an exact chain score and, for near-ties, base-level score differences computed only where the candidates differ.
Fast and small
26 to 83 times faster than minimap2 on the transcriptome, in at most 2.1 GB of memory, with a 1.3 GB index for all of GENCODE.
Made for quantification
The scored transcript sets are what an EM needs: oarfish quantifies spearfish’s RAD output, or uses spearfish as its mapping front end.