Comprehensive benchmarking of methods for mutation calling in circulating tumor DNA

Area of Science:

• Genomics
• Cancer Research
• Bioinformatics

Background:

• Detecting somatic mutations in cell-free DNA (cfDNA) is limited by low variant allele frequencies and DNA degradation.
• Accurate detection of tumor-derived DNA is crucial for non-invasive cancer diagnostics and monitoring.

Purpose of the Study:

• To develop a robust benchmarking strategy for evaluating somatic variant callers in cfDNA.
• To assess the performance of variant callers across various circulating tumor DNA (ctDNA) levels and sequencing depths.
• To provide guidance for selecting optimal variant calling methods for liquid biopsy applications.

Main Methods:

• Utilized longitudinal, patient-matched cfDNA samples from colorectal and breast cancer patients.
• Created controlled dilution series by combining samples with high and ultra-low tumor-derived DNA levels.
• Performed deep whole-genome (150x) and exome (2,000x) sequencing to identify variants.
• Benchmarked nine somatic variant callers and explored machine learning for caller optimization.

Main Results:

• Defined a reference set of approximately 37,000 single nucleotide variants and 58,000 indels.
• Clarified the detection limits of current somatic variant calling approaches in cfDNA.
• Identified key features that enhance variant caller accuracy for cfDNA analysis.

Conclusions:

• The developed benchmarking strategy provides a reliable resource for assessing cfDNA analysis methods.
• The findings offer practical guidance for selecting and optimizing somatic variant callers in liquid biopsy.
• Improved variant detection in cfDNA has significant implications for cancer diagnostics and personalized medicine.