Rawsamble: overlapping raw nanopore signals using a hash-based seeding mechanism

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Raw nanopore signal analysis offers fast, resource-efficient genomics without basecalling.
• Existing methods struggle with noisy raw signal comparisons for unknown genomes.
• Direct analysis of raw signals without a reference genome is a key challenge.

Purpose of the Study:

• To enable direct analysis of raw nanopore signals without a reference genome.
• To develop a mechanism for identifying similarity between all raw signal pairs (all-vs-all overlapping).

Main Methods:

• Proposed Rawsamble, a novel hash-based search mechanism for all-vs-all raw signal overlapping.
• Utilized Rawsamble overlaps with the miniasm assembler for de novo assembly graph construction.
• Evaluated performance across multiple genomes of varying sizes.

Main Results:

• Achieved significant speedup (5.01× on average, up to 23.10×) and reduced peak memory usage (5.74× on average, up to 22.00×) compared to conventional pipelines.
• Constructed de novo assemblies directly from raw signals, a first in the field.
• Generated accurate unitigs up to 2.3 million bases long, comparable to state-of-the-art methods.

Conclusions:

• Rawsamble facilitates efficient de novo genome assembly directly from raw nanopore signals.
• The method offers substantial computational advantages over traditional basecalling-dependent pipelines.
• Rawsamble represents a significant advancement for reference-free genomics analysis.