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

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• PacBio High-Fidelity (HiFi) sequencing generates highly accurate long reads (>99%).
• De novo sequence assemblers utilize error correction (EC) as a primary step for processing HiFi data.
• The performance of EC algorithms on this new data type remains unevaluated.

Purpose of the Study:

• To introduce hifieval, a novel command-line tool for quantifying over- and under-corrections in EC algorithms.
• To assess the accuracy of EC components within existing HiFi assemblers.
• To investigate EC performance in challenging genomic regions.

Main Methods:

• Development of the hifieval tool for measuring EC accuracy.
• Evaluation of EC components from current HiFi assemblers.
• Testing on CHM13 and HG002 datasets, including analysis of homopolymer, centromeric, and segmental duplication regions.

Main Results:

• The study provides the first evaluation of EC accuracy for HiFi sequencing data.
• Performance variations in EC methods were identified across different genomic contexts.
• hifieval enables quantitative assessment of EC over- and under-corrections.

Conclusions:

• hifieval is a valuable tool for benchmarking and improving EC algorithms.
• Enhanced EC accuracy is crucial for advancing de novo assembly quality with HiFi data.
• This work will guide the development of more robust HiFi assemblers.