| 23 |
2026 |
|
American Assoiation for Cancer Research (AACR) |
リンクを見る
|
| 22 |
2026 |
|
American Assoiation for Cancer Research (AACR) |
リンクを見る
|
| 21 |
2026 |
|
2026 ASCO Gastrointestinal Cancers Symposium |
リンクを見る
|
| 20 |
2026 |
Deep learning–based cfDNA fragment–level analysis: A genome-scanning approach for detection of colorectal cancer.
Abstract
Background:Cell-free DNA (cfDNA) has emerged as a promising biomarker for cancer detection. Circulating tumor DNA (ctDNA) contains distinct genomic and epigenomic features, including mutations and fragmentomic patterns, that differentiate cancer patients fr|om healthy individuals. This study aimed to develop a deep-learning–based cancer detection method by integrating fragment-level genomic and epigenomic features.
Methods:The study analyzed cfDNA samples fr|om 768 healthy individuals and 150 colorectal cancer patients (stages I–IV). A deep-learning framework based on a 2D convolutional neural network (CNN) was developed using multi-channel genomic tensors derived fr|om sequencing reads. Six strand-specific channels, including nucleotide sequences, methylation status, and fragment alignment characteristics, were integrated for analysis. Model performance was compared with conventional approaches such as Average Methylation Fraction (AMF), Depth Profile, Fragment Size Ratio (FSR), and end-motif analysis.
Results:The proposed method demonstrated superior performance compared with conventional methods, achieving the highest mean per-marker AUC (0.769 ± 0.039). The final model achieved an overall AUC of 0.905 in the independent test set. The model also showed strong performance for both early-stage (I–II) colorectal cancer detection (AUC 0.880) and advanced-stage (III–IV) disease detection (AUC 0.929).
Conclusions:Integrating fragment-level cfDNA genomic and epigenomic features enables highly accurate colorectal cancer detection, including early-stage disease. The genome-scanning deep-learning approach also improves interpretability by visualizing cancer-associated genomic regions, highlighting its potential as a precise and clinically useful liquid biopsy platform.
|
2026 ASCO Gastrointestinal Cancers Symposium |
リンクを見る
|
| 19 |
2026 |
Preliminary results of tumor-informed minimal residual disease detection in pancreatic cancer patients following curative resection and FOLFIRINOX adjuvant chemotherapy.
Abstract
Background:Circulating tumor DNA (ctDNA)-based minimal residual disease (MRD) testing has emerged as a promising prognostic tool in solid tumors and may help guide postoperative treatment strategies. This study evaluated the prognostic significance of ctDNA MRD monitoring in patients with pancreatic cancer who underwent curative resection followed by adjuvant FOLFIRINOX therapy.
Methods:Between October 2023 and February 2025, patients with resected pancreatic cancer receiving adjuvant FOLFIRINOX were prospectively enrolled fr|om 11 hospitals in Korea. Blood samples were collected at up to seven postoperative time points and analyzed using a tumor-informed ctDNA assay (CancerDetect, IMBdx). MRD positivity was defined as the detection of two or more mutations.
Results:A total of 86 patients were included in the analysis. The postoperative MRD positivity rate at the initial time point (P1) was 31.3%, and positivity increased with advancing disease stage. During a median follow-up of 13 months, 23.3% of patients experienced recurrence. Conventional clinicopathologic factors, including stage and T/N classification, were not significantly associated with disease-free survival (DFS). In contrast, MRD positivity at P1 was strongly associated with inferior DFS (HR 5.31, P<0.001). The prognostic impact of MRD was particularly significant in stage I and II disease. Longitudinal MRD analysis further demonstrated that patients with persistent MRD positivity had the poorest outcomes, while dynamic MRD changes were closely associated with treatment response and recurrence risk.
Conclusions:Postoperative ctDNA MRD detection strongly predicted recurrence in resected pancreatic cancer treated with adjuvant FOLFIRINOX. Longitudinal MRD dynamics also correlated with treatment response and recurrence risk, supporting the clinical utility of ctDNA MRD for postoperative surveillance and therapeutic decision-making.
|
2026 ASCO Gastrointestinal Cancers Symposium |
リンクを見る
|
