Bi-level error correction for PacBio long reads.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Next-generation sequencing technologies like PacBio and Oxford Nanopore produce long reads but suffer from high error rates (up to 15%).
• Accurate long reads are crucial for complex genomic analyses, but existing error correction methods face challenges with noisy data.
• Hybrid error correction strategies, leveraging low-error Illumina reads, show promise for improving long-read quality.

Purpose of the Study:

• To introduce Bicolor, a novel bi-level hybrid error correction framework for enhancing PacBio long-read sequencing data.
• To improve the accuracy, identity ratio, and alignment performance of long reads for downstream genomic applications.

Main Methods:

• Bicolor employs a two-level hybrid error correction strategy.
• Level 1 utilizes a de Bruijn graph approach with iterative solid k-mer path searching.
• Level 2 integrates results from Level 1 using multiple sequence alignment and a voting algorithm for final base determination.

Main Results:

• Bicolor consistently achieved the highest identity ratio across three real datasets.
• The method demonstrated superior alignment ratios and a higher number of aligned reads compared to state-of-the-art methods on two datasets.
• Bicolor showed competitive performance in alignment and genome coverage on the third dataset.

Conclusions:

• Bicolor offers a significant advancement in hybrid error correction for PacBio long reads.
• The proposed method effectively improves sequencing data quality, enabling more reliable genomic analyses.
• Bicolor provides a robust and high-performing solution for error correction in long-read sequencing data.