The problem. Short-read aligners like BWA were built for accurate ~100 bp reads. Nanopore and PacBio produce reads tens of kilobases long with much higher error rates, and the alignment problem changes shape: you need to seed and chain across long, error-riddled sequences, handle large structural gaps, and do it fast enough for whole genomes. The old tools didn’t fit.

The idea. Minimap2 seeds alignments with minimizers, chains them with a concave gap cost that tolerates long indels, and only then fills in base-level alignment where needed. One code path handles long genomic reads, spliced long-read mRNA, short reads, and genome-to-genome alignment — configured by preset rather than rewritten. It’s several times faster than short-read mappers and dozens of times faster than earlier long-read tools at comparable or better accuracy.

Why it matters. This is the aligner underneath the long-read world I read about earlier — the Nanopore and PacBio HiFi papers, and the assemblers that consume its output. It’s the natural successor to BWA in my own variant-calling instincts: same author, same minimizer-and-chain lineage, extended to the reads that break short-read tools. If you touch long reads, this is the first step.

Verdict. The de facto standard for long-read alignment; presets matter, and base-level accuracy still depends on read quality. Read it as the workhorse that turned long, noisy reads into usable alignments.