Cancer signature ensemble integrating cfDNA methylation, copy number, and fragmentomic patterns for multi-cancer early detection.

Abstract

e15040 Background: The use of cfDNA sequencing has demonstrated great potential for cancer screening; particularly, methylation signatures, copy number variations, and fragmentomic profiles have proven to be effective for identifying early cancer signals. However, most large-scale studies have only focused on either targeted methylation sites or whole-genome sequencing, limiting comprehensive analysis that integrates both epigenetic and genetic signatures. In this study, we present a platform for multi-cancer early detection that enables simultaneous analysis of whole-genome methylation, copy number, and fragmentomic patterns of cfDNA in a single assay. Methods: For a total of 950 plasma samples (361 healthy controls and 107 colon, 113 liver, 238 lung, and 131 prostate cancer) and 239 tissue samples, whole-genome methylation sequencing data were generated. Machine learning was conducted for multiple feature types engineered from cfDNA samples, and independent test performance was assessed. Results: A multi-feature cancer signature ensemble (CSE) classifier, integrating all features, outperformed single-feature classifiers. At 95.2% specificity, the cancer detection sensitivity with methylation, copy number, and fragmentomic models were 77.2% [CI 72-82], 61.4% [CI 56-67], and 60.5% [CI 55-66], respectively; but it was significantly increased to 87.7% [CI 84-91] with CSE ( p-value < 0.0001). For tracing the tissue of origin, CSE enhanced the accuracy beyond the methylation classifier, from 74.7% [CI 69-80] to 77.5% [CI 72-82]. Conclusions: This work proves the utility of a signature ensemble integrating epigenetic and genetic information for accurate cancer detection.