Sović, I. et al. Nat. Commun. 7, 11307 (2016).

The recent availability of the first protein-nanopore-based commercial sequencer has enabled the production of very long reads from single-molecule DNA templates. Taking advantage of this output is made complicated by the high random error rate in individual reads and the variability in quality and read lengths. Sović et al. introduce GraphMap, an efficient computational tool for mapping long sequence reads that is largely robust to error profile and error rate. On benchmark synthetic data modeled on a number of organisms, including human, GraphMap shows sensitivity similar to that of the BLAST gold standard, but it is several orders of magnitude faster. The method also performs well on real data relative to competing tools, and it enables downstream applications such as pathogen detection and robust calling of single-nucleotide polymorphisms and structural variants.