Research analyzing cell-free DNA fragmentation patterns in healthy individuals reveals physiological confounders that could improve test accuracy.
GC Genome published research identifying physiological factors that can interfere with cancer-associated cell-free DNA signals in liquid biopsy tests, potentially improving the accuracy of multi-cancer early detection assays.
The study, published in Clinical Chemistry, analyzed cfDNA fragmentation patterns in 1,154 healthy individuals to identify potential confounders that could influence cfDNA-based cancer detection. The research was conducted in collaboration with professor Min-Jung Kwon and her team at Kangbuk Samsung Medical Center.
Researchers examined correlations between cfDNA fragmentomic profiles and 65 clinical variables, including age and liver function markers. The goal was to identify non-cancer physiological factors that could produce false signals in individuals without cancer.
Key Findings:
- Liver enzymes(including AST, ALP, γ-GTP) and age were identified as major factors altering cfDNA fragmentation patterns.
- Elevated AST or age closely resembled cancer-like fragmentomic signatures, blurring the distinction between noncancer and cancer profiles.
- AST showed high similarity to fragmentation size patterns seen in lung cancer patients (cosine similarity = 0.98).
- Age showed the highest similarity to cancer-like profiles among clinical variables (cosine similarity = 0.52).
- Receiver Operating Characteristic (ROC) analysis confirmed that these physiological variables can act as confounders by reducing the specificity of cfDNA-based detection, potentially leading to false-positive results.
The findings demonstrate that various physiological factors can influence cfDNA signals, highlighting the need for confounder-aware modeling approaches in liquid biopsy development.
“This study is significant because it uses large-scale data from healthy individuals to identify key confounders that influence cfDNA fragmentation patterns,” says a GC Genome spokesperson in a release. “These insights will play an important role in refining our Multi-Cancer Early Detection test, ai-CANCERCH, particularly in reducing false-positive rates and improving test specificity.”
Implications for Multi-Cancer Detection
The research supports improvements to GC Genome’s ai-CANCERCH test, an AI-based multi-cancer early detection assay that uses low-coverage whole-genome sequencing. The test detects signals associated with multiple cancers using 10 mL of blood.
GC Genome plans to expand ai-CANCERCH from detecting six cancers to 10 cancers in January 2026. The expanded panel will include colorectal, lung, esophageal, liver, ovarian, pancreatic, biliary, breast, gastric, and head-and-neck cancers.
The identification of physiological confounders addresses a key challenge in liquid biopsy development, where non-cancer factors can generate signals that mimic cancer-associated changes in cfDNA. Understanding these confounders allows developers to build more specific algorithms that can distinguish between cancer signals and physiological noise.
ID 118203373 © Oleg Dudko | Dreamstime.com
