Abstract 170P
Background
Gastric cancer ranks among the leading causes of cancer-related mortality worldwide. Early detection is essential for effective treatment and improved survival outcomes. Circulating tumor DNA (ctDNA) released by tumors into the bloodstream has emerged as a promising non-invasive biomarker for the early detection of gastric cancer. However, plasma cell-free DNA (cfDNA) from patients with benign lesions, such as gastritis, often exhibits overlapping molecular signatures with those from patients with early-stage gastric cancer, posing a significant challenge in distinguishing cancer-specific signatures. In this study, we developed a multimodal approach to identify multiple signatures in plasma that facilitates the development of a blood test capable of accurately differentiating gastric cancer patients from individuals with gastritis.
Methods
We recruited fifty-three patients with early-stage gastric cancer (stages I-III) and thirty-five patients with gastritis. Using a multimodal approach, we simultaneously profiled methylation, copy number alterations (CNA), end motifs, and fragment length in plasma samples from these patients. A machine learning model was constructed, utilizing the important signatures as input data, to classify gastric cancer patients from those with gastritis.
Results
We observed genome-wide hypomethylation, an enrichment of long fragments (>150 bp), and a reduced frequency of thymine at cleavage sites in cfDNA isolated from gastric cancer patients compared to those with gastritis. Our machine learning model achieved an AUC of 86.11%, a sensitivity of 94.44%, and a specificity of 81.82% in distinguishing these two groups of patients.
Conclusions
The multimodal approach based on methylation and fragment patterns of cfDNA has demonstrated high accuracy in the early detection of gastric cancer, reducing false positive results caused by gastritis lesions. These findings require prospective validation in a larger cohort.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
Gene Solutions JSC.
Disclosure
All authors have declared no conflicts of interest.