Curated variation benchmarks for challenging medically relevant autosomal genes

Area of Science:

• Genomics
• Clinical Genetics

Background:

• Medically relevant genes often present repetitive or polymorphic complexity, challenging accurate clinical analysis.
• Existing variant benchmark sets exclude nearly 400 such genes, limiting comprehensive genomic assessment.

Purpose of the Study:

• To characterize 273 challenging autosomal genes.
• To develop an improved benchmark set for medically relevant genes.
• To address limitations in current variant detection technologies.

Main Methods:

• Haplotype-resolved whole-genome assembly was used to characterize 273 challenging autosomal genes.
• A curated benchmark was generated for human genome references GRCh37 and GRCh38.
• Variant calling was performed using short- and long-read technologies.

Main Results:

• Over 17,000 single-nucleotide variations, 3,600 insertions/deletions, and 200 structural variations were reported for each reference genome.
• False duplications in GRCh37/GRCh38 led to missed variants in key genes (e.g., CBS, CRYAA, KCNE1).
• Masking false duplications improved variant recall from 8% to 100%.

Conclusions:

• Haplotype-resolved assembly provides a robust method for creating comprehensive variant benchmarks.
• Addressing reference genome inaccuracies is crucial for accurate clinical variant detection.
• This approach can serve as a prototype for future whole-genome benchmark development.