Nanopore sequencing is able to sequence very long reads, but sequence assembly remains necessary to reconstruct full genomes and perform analyses without references. To deal with the large amounts of data and real-time constraints, various methods have been proposed to avoid computational bottlenecks by working directly with the raw nanopore electrical signals emitted during sequencing. However, the majority of such methods are focused on reference-based analyses such as mapping.
Rawsamble is the first work to demonstrate assembly directly from raw nanopore electrical signals using overlaps to construct a string-graph. Constructing such graphs comes with its own set of computational challenges, which can be avoided using de Bruijn graphs. At the same time, de Bruijn methods suffer from error sensitivity, a considerable challenge when working with noisy raw nanopore electrical signals. Here we explore de Bruijn-like methods for assembly, using fuzzy de Bruijn graphs to improve error tolerance by allowing inexact matches between signals. On real datasets, we achieve higher contiguity of the assembly while matching the performance advantages of Rawsamble.
Zoom Link: https://umd.zoom.us/j/