spearfish: ultrafast Δ-mapping of long reads to redundant transcriptomes

spearfish’s Δ-mapping for long-read transcriptomes computes alignment-score differences while skipping shared sequence, enabling 26–83x faster mapping than minimap2 with up to 1/7 memory.

• Why it matters: Isoforms and gene-family near-copies create redundant targets, but common long-read mappers repeatedly seed, chain, and align each isoform separately, wasting time and memory.
• What they did: spearfish builds a core graph from locally consistent parsing (LCP) and computes exact colinear chaining scores over it, sharing dynamic-programming state across transcripts.
• The result: Across RNA-seq experiments (32 threads), it is 2.6–11.3x faster than lr-kallisto pseudoalignment and, with oarfish, yields abundance estimates matching or improving minimap2-based workflows while quantification is 3.0–8.8x faster.

AI summary of the abstract · may contain errors · not medical advice · read the paper.