The problem. SNVs and small indels are what GATK and DeepVariant call well from short reads. But a large fraction of consequential genetic variation is structural: kilobase deletions, inversions, translocations, and duplications. These span more than a short read, so they leave only indirect, ambiguous signatures in short-read data — exactly the class of variation day 9’s nanopore papers existed to reach.

The idea. Sniffles detects structural variants from long single-molecule reads (nanopore, PacBio). Because a long read can span an entire rearrangement, the SV appears directly within individual reads as split or clipped alignments; Sniffles aggregates these signals across reads to call complex and nested events accurately, with breakpoints.

Why it matters. This is the missing half of variant calling relative to variant_calling_nf. My pipeline is short-read and small-variant by design — reading Sniffles marks exactly where that design stops and what a long-read complement would add. It closes the loop from day 9: the long reads that assemble hard regions are also what let you call the structural variants short reads can’t.

Verdict. Foundational for long-read structural-variant calling and widely used. Its reach is bounded by long-read availability and error profiles, but for SVs it sees what short reads cannot. Read it as the structural-variation counterpart to the small-variant callers.