The problem. Small-variant callers stop at substitutions and short indels. Larger events, such as deletions, duplications, inversions, and translocations, leave a different fingerprint: read pairs that map too far apart, and reads that split across a break. Detecting these from short reads is slow if you check every candidate carefully.

The idea. Manta scans the genome for the paired-end and split-read signals that mark a breakpoint, builds a short list of candidate regions, and then locally assembles each one to pin the break to the base. It is engineered for speed, so a 50x genome finishes in under half an hour on standard hardware. The same run covers both germline samples and tumor-normal pairs, and it reports mid-size indels that small-variant callers miss.

Why it matters. Structural variants are the gap between my small-variant calling and the full picture of a genome, and they matter in AML, where fusions and large rearrangements drive disease. Manta is the short-read counterpart to Sniffles (day 16), which reads the same events off long reads. Running both would let me cross-check rearrangements from two very different data types.

Verdict. A fast, well-built SV caller that became a common default. Short reads limit resolution in repeats, where long-read methods do better. Read it as the quick way to pull structural variants out of the data I already have.