Adaptively integrated sequencing and assembly of near-complete genomes

Recent advances in long-read sequencing (LRS) and assembly algorithms have made it possible to generate highly complete genome assemblies for humans, animals, plants and other eukaryotes. However, there is a need for ongoing development to improve the accessibility and affordability of the required data, increase the range of usable sample types, and reliably resolve the most challenging, repetitive genome regions. Here we introduce 'Cornetto', a new experimental paradigm in which the genome assembly process is adaptively coupled to Oxford Nanopore's programmable selective sequencing functionality, with target regions being iteratively updated to focus LRS data production onto the unsolved regions of a nascent assembly. Cornetto improves assembly quality and streamlines the process, both for human individuals and non-human vertebrates tested here. Cornetto enables us to generate highly complete diploid human genome assemblies using only a single LRS platform, surpassing the quality of previous efforts at a fraction of the cost. We use Cornetto to generate similar assemblies from challenging sample types like human saliva, for the first time, thereby further enhancing accessibility. Finally, we obtain complete and accurate assemblies for clinically-relevant repetitive loci at the extremes of the genome, demonstrating valid approaches for genetic testing in Facioscapulohumeral muscular dystrophy (FSHD) and Autosomal Dominant Tubulointerstitial Kidney Disease (ADTKD) – inherited diseases for which diagnosis is complicated by an inability to sequence the genes involved. In summary, Cornetto will improve, accelerate and democratise genome assembly, with broad impacts across bioscience domains.