Benchmarking UMI-aware and standard variant callers for low frequency ctDNA variant detection

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• Circulating tumor DNA (ctDNA) is a promising biomarker for cancer treatment decisions.
• Accurate detection of low-frequency ctDNA variants is challenging due to sequencing and PCR artifacts.
• Unique Molecular Identifiers (UMIs) aid in filtering artifacts from sequencing data.

Purpose of the Study:

• To benchmark the performance of UMI-aware variant calling tools against standard tools for ctDNA analysis.
• To evaluate the ability of different variant callers to accurately detect low-frequency ctDNA variants.

Main Methods:

• Benchmarking six variant calling tools: Mutect2, bcftools, LoFreq, FreeBayes, UMI-VarCal, and UMIErrorCorrect.
• Utilizing datasets with known low-frequency variants and actual ctDNA samples.
• Generating synthetic UMI sequences to assess artifact filtering capabilities.

Main Results:

• Variant callers exhibited varying sensitivity and specificity profiles.
• Mutect2 demonstrated high sensitivity but also a higher rate of false positives without UMIs.
• UMI-VarCal showed superior performance in reducing false positives in UMI-encoded data.
• Mutect2 achieved a balance of sensitivity and specificity with synthetic UMIs.

Conclusions:

• UMI-aware variant callers offer improved sensitivity and specificity for low-frequency ctDNA variant detection compared to standard tools.
• Further development of UMI-aware tools is crucial for realizing effective early cancer detection using ctDNA.