Somatic variant-calling beyond cancer: Repurposing algorithms to map low‑allele‑fraction variants across genomics

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Somatic variant callers were designed for tumor-normal comparisons.
• Non-cancer fields like developmental biology and virology increasingly need low-frequency variant detection.
• Existing cancer callers are often repurposed without clear validation for these new applications.

Purpose of the Study:

• To review somatic variant callers and classifiers for non-cancer, low-allele-fraction variant detection.
• To synthesize case studies (2020-2025) applying these tools outside oncology.
• To identify challenges and propose adaptations for repurposing somatic callers.

Main Methods:

• Review of widely used somatic callers and post-calling classifiers.
• Synthesis of peer-reviewed case studies (2020-2025) using these tools for low-allele-fraction variants.
• Extraction of data on sample type, sequencing depth, allele fraction thresholds, callers, and validation.

Main Results:

• Cancer callers typically assume diploidy and moderate allele fractions, using Beta-binomial or Poisson models.
• Non-cancer applications often involve allele fractions below 5%, complicated by ploidy and sequencing artifacts.
• Successful repurposing demonstrated for post-zygotic variants (1-3%), sperm mosaicism, mitochondrial heteroplasmy (0.4-0.5%), viral variants (~5%), and cell-free DNA (cfDNA) variants (≥1-5%).
• Techniques like unique molecular identifiers, panel-of-normal filtering, and machine learning enhance specificity.
• Emerging long-read and deep-learning callers show promise for improved sensitivity.

Conclusions:

• Somatic callers provide a versatile framework for interrogating low-allele-fraction variants.
• Careful parameter tuning and validation are essential for reliable results in diverse biological settings.
• Future integration of pangenome references and federated benchmarking can enhance robustness.