Related Experiment Video
Updated: Jun 7, 2025
11:02
Detecting Somatic Genetic Alterations in Tumor Specimens by Exon Capture and Massively Parallel Sequencing
Published on: October 18, 2013
19.4K
Evaluating Bioinformatics Processing of Somatic Variant Detection in cfDNA Using Targeted Sequencing with UMIs
Yixin Lin1,2, Mads Heilskov Rasmussen1,2, Mikkel Hovden Christensen1,2
1Department of Molecular Medicine, Aarhus University Hospital, 8200 Aarhus, Denmark.
International Journal of Molecular Sciences
|November 9, 2024
Summary
Accurately detecting cancer DNA (ctDNA) in blood is hard. This study found shearwater-AND best for tumor mutations and DREAMS-vc for cancer detection, depending on the analysis type.
Area of Science:
- Genomics
- Molecular Biology
- Cancer Research
Background:
- Circulating tumor DNA (ctDNA) shows promise as a cancer biomarker.
- Detecting low-frequency tumor mutations in cell-free DNA (cfDNA) is difficult due to low variant allele frequencies and sequencing errors.
Purpose of the Study:
- To benchmark variant callers (Mutect2, VarScan2, shearwater, DREAMS-vc) for cfDNA analysis.
- To evaluate performance at both mutation and sample classification levels.
- To investigate the impact of Unique Molecular Identifiers (UMIs) and consensus strategies.
Main Methods:
- Deep targeted sequencing of cfDNA with UMIs from 111 colorectal cancer patients.
- Benchmarking of four variant callers: Mutect2, VarScan2, shearwater, and DREAMS-vc.
- Assessment of mutation-level precision and sample-level classification accuracy (ROC-AUC).
- Evaluation of UMI grouping and consensus strategies, including network-based methods.
- Analysis of sequencing depth effects and downsampling strategies.
Main Results:
- The shearwater-AND method achieved the highest precision for detecting tumor mutations.
- shearwater-AND reached an ROC-AUC of 0.984 for sample classification in tumor-informed analyses.
- DREAMS-vc achieved the highest ROC-AUC of 0.808 for sample classification in tumor-agnostic studies.
- Sequencing depth variations in PBMCs can cause false positives, mitigated by downsampling.
- Network-based UMI grouping improved performance compared to identical UMI grouping.
Conclusions:
- The optimal cfDNA variant caller is context-dependent, varying with study goals (mutation vs. sample classification) and approach (tumor-informed vs. tumor-agnostic).
- shearwater-AND is highly effective for tumor-informed mutation detection and classification.
- DREAMS-vc excels in tumor-agnostic sample classification.
- Careful consideration of UMI strategies and sequencing depth is crucial for accurate cfDNA analysis.
