Related Experiment Video
Updated: Oct 6, 2025
06:53
Detection and Monitoring of Tumor Associated Circulating DNA in Patient Biofluids
Published on: June 8, 2019
8.8K
Error Characterization and Statistical Modeling Improves Circulating Tumor DNA Detection by Droplet Digital PCR
Tenna V Henriksen1,2, Simon O Drue1,2, Amanda Frydendahl1,2
1Department of Molecular Medicine, Aarhus University Hospital, Aarhus, Denmark.
Clinical Chemistry
|January 14, 2022
Summary
Accurate error modeling is crucial for sensitive circulating tumor DNA (ctDNA) detection using droplet digital PCR (ddPCR). New noise-informed methods, including CASTLE, improve ctDNA calling accuracy, especially for transition assays and varying DNA input levels.
Area of Science:
- Molecular Biology
- Genomics
- Cancer Research
Background:
- Droplet digital PCR (ddPCR) is a sensitive method for detecting circulating tumor DNA (ctDNA).
- Low abundance of ctDNA necessitates robust methods to distinguish true signals from noise.
- Characterizing ddPCR-generated noise is essential for improving ctDNA detection sensitivity and specificity.
Purpose of the Study:
- To characterize ddPCR-generated noise for mutation-detecting assays.
- To develop novel, noise-informed ctDNA calling methods.
- To compare the performance of new methods against existing literature-established approaches.
Main Methods:
- Empirically characterized noise profiles for 70 mutation-detecting ddPCR assays using 95 nonmutated samples.
- Developed two novel calling methods: dynamic limit of blank and concentration and assay-specific tumor load estimator (CASTLE).
- Assessed method performance on 9447 reference samples and 1311 colorectal cancer patient plasma samples, comparing against 7 established methods.
Main Results:
- Assay noise often correlated with DNA input amount; transition assays were more error-prone than transversion assays.
- Both novel calling methods effectively accounted for assay-specific noise, maintaining high performance across varying DNA inputs.
- Noise-uninformed methods showed lower performance; CASTLE uniquely provided noise-corrected mutation level and call certainty estimates.
Conclusions:
- Accurate error modeling is fundamental for sensitive and specific ctDNA detection via ddPCR.
- Accounting for DNA input concentration enhances detection specificity.
- CASTLE is presented as a robust tool for ctDNA calling in ddPCR applications.
