Evaluation of haplotype-aware long-read error correction with hifieval.

Area of Science:

• Genomics
• Bioinformatics

Background:

• PacBio High-Fidelity (HiFi) sequencing offers long reads with >99% accuracy, enabling advanced *de novo* genome assembly.
• Current HiFi assemblers rely on sequencing error correction as a primary step, but this critical component lacks prior evaluation for this new data type.

Approach:

• Introduced hifieval, a novel command-line tool designed to quantify over- and under-corrections in sequence error correction algorithms.
• Assessed the accuracy of error correction modules within existing HiFi assemblers using the CHM13 and HG002 datasets.
• Investigated error correction performance in challenging genomic regions, including homopolymers, centromeres, and segmental duplications.

Key Points:

• Hifieval provides a standardized method for evaluating error correction accuracy in HiFi sequencing data.
• Performance varied across assemblers, particularly in complex genomic areas.
• Identified specific challenges in homopolymer regions, centromeric DNA, and segmental duplications.

Conclusions:

• Hifieval is crucial for benchmarking and improving the error correction step in HiFi sequence assemblers.
• The tool will facilitate the development of more robust assemblers, leading to higher quality genome assemblies.
• Future work can leverage hifieval to refine algorithms and enhance genome assembly accuracy for diverse genomic contexts.