Benchmarking Low-Frequency Variant Calling With Long-Read Data on Mitochondrial DNA.

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Long-read sequencing technologies have advanced, improving the detection of low-frequency somatic variants.
• Mitochondrial DNA analysis is crucial for understanding various diseases and evolutionary processes.
• Assessing the accuracy of nanopore sequencing for detecting low-frequency variants is essential for its clinical application.

Purpose of the Study:

• To benchmark nanopore sequencing for low-frequency single nucleotide variant detection in mitochondrial DNA.
• To evaluate the impact of different base-calling, alignment, and variant-calling tools on accuracy.
• To compare nanopore sequencing performance against a short-read gold standard.

Main Methods:

• Utilized mixtures of mitochondrial samples with known haplogroups at 5%, 2%, and 1% levels.
• Employed two base-callers (Guppy versions), two aligners (Minimap2, Ngmlr), and three variant callers (Mutserve2, Freebayes, Nanopanel2).
• Assessed variant-calling performance using F1 scores and compared results to a short-read gold standard.

Main Results:

• Nanopore sequencing yielded a mean read length of 11 kb and mean quality of 15.
• Ngmlr aligner showed higher F1 scores but also higher false-positive allele frequencies compared to Minimap2.
• Mutserve2 variant caller achieved the highest F1 scores across all mixture levels, particularly excelling at lower frequencies.

Conclusions:

• Nanopore sequencing is a viable tool for low-frequency mitochondrial variant detection.
• Mutserve2 demonstrates superior performance for variant calling in this context.
• This study identifies current limitations and provides a benchmark for optimizing nanopore-based variant detection pipelines.