Long-read amplicon denoising.

Area of Science:

• Genomics
• Bioinformatics

Background:

• Long-read sequencing technologies offer potential for comprehensive gene and genome analysis.
• However, inherent error profiles in long reads complicate data processing and downstream applications.

Purpose of the Study:

• To develop accurate methods for reconstructing error-free DNA sequences and their frequencies from PacBio long reads.
• To address the limitations of existing amplicon denoising techniques for high indel error rates in long reads.

Main Methods:

• Introduced two novel amplicon denoising algorithms: Fast Amplicon Denoising (FAD) and Robust Amplicon Denoising (RAD).
• FAD is optimized for speed and accuracy with medium-length reads and high coverage.
• RAD provides robustness for very long reads or lower coverage scenarios.

Main Results:

• Both FAD and RAD demonstrated superior accuracy compared to existing methods on simulated and Mock Virus Community datasets.
• The methods successfully identified sequence variants differing by a single nucleotide.
• Competitive run times were achieved, even for the more robust RAD method.

Conclusions:

• FAD and RAD effectively address the challenge of amplicon denoising for long-read sequencing data.
• These methods improve the accuracy and reliability of variant reconstruction from complex populations.
• Available Julia implementations and a webserver facilitate broader adoption and application.