Researchers have released SVPG, software that finds large DNA changes and adds them to a shared genome reference without rebuilding that reference from scratch. In a benchmark using 20 samples, its update pipeline ran nearly ten times faster than an approach that first reconstructs each sample's genome.
A conventional reference represents DNA as one long sequence. That makes comparison difficult when a person's DNA contains an extra, missing or rearranged stretch. A pangenome graph stores alternative sequences as branches, so researchers can compare a new sample with variation already observed in other people.
SVPG uses that graph in two ways. Its guided mode revisits potential structural variants, changes larger than 50 DNA base pairs, first detected against a conventional reference. It compares the relevant sequencing reads, the DNA fragments recorded by a sequencing machine, with the graph to refine the locations of the changes and remove false detections. Its graph mode begins with reads already matched to the graph and identifies variants that the existing paths do not contain.
Those new variants can then be added to the graph. This is the step behind the faster 20-sample update: the pipeline inserts the detected changes into an existing reference instead of first assembling a complete genome from every new sample. It connects finding variation with maintaining the reference that later samples will use.
SVPG also tested well on a standard Genome in a Bottle sample. In high-confidence regions, its guided mode scored 0.958 with Oxford Nanopore data and 0.966 with PacBio HiFi data. These F1 scores balance how many real variants the tool finds against how many reported variants are correct. The scores describe those sequencing methods and benchmark regions, not diagnostic accuracy across all patients or all parts of a genome.
Repeated DNA regions remain difficult, and mistakes in the reference graph or in matching reads to it can affect the result. The authors have released code, benchmark outputs and update scripts, allowing researchers to examine that trade-off. The demonstrated speed gain makes incremental reference updates more practical for the tested samples; the study does not establish how the same balance of speed and accuracy scales to thousands of genomes.