Scalable long read self-correction and assembly polishing with multiple sequence alignment.

Area of Science:

• Genomics
• Bioinformatics

Background:

• Third-generation sequencing offers long reads but has high error rates (~10%).
• Self-correction is crucial for analyzing long sequencing reads.
• Existing methods struggle with scalability for ultra-long reads.

Purpose of the Study:

• Introduce CONSENT, a novel self-correction method for long sequencing reads.
• Evaluate CONSENT's performance, scalability, and impact on genome assembly.
• Develop an efficient method for processing ultra-long reads.

Main Methods:

• Developed CONSENT, combining multiple sequence alignment and local de Bruijn graphs.
• Implemented an efficient segmentation strategy for scalable multiple sequence alignment.
• Tested CONSENT on Oxford Nanopore data, including ultra-long reads up to 1.5 Mbp.

Main Results:

• CONSENT demonstrates superior performance on real Oxford Nanopore data compared to state-of-the-art methods.
• It uniquely scales to process ultra-long reads, handling a human dataset in 10 days.
• Error correction with CONSENT improved Flye assemblies; polishing feature showed 2-38x speedup.

Conclusions:

• CONSENT provides an efficient and scalable solution for long-read error correction.
• Directly assembling raw data and polishing the assembly is more efficient and yields better results than read correction followed by assembly.
• CONSENT offers a valuable tool for improving genome assembly quality from long-read sequencing data.