Nguyen, T. H. H. | bioRxiv Science

bioRxiv2024

Supervised non-negative matrix factorization on cell-free DNA fragmentomic features enhances early cancer detection

BackgroundCell-free circulating DNA (cfDNA) fragments exhibit non-random patterns in their length (FLEN), end-motif (EM), and distance to nucleosome position (ND). While these cfDNA features have shown promise as inputs for machine learning and deep learning models in early cancer detection, most studies utilize them as raw inputs, overlooking the potential benefits of pre-processing to extract cancer-specific features. This study aims to enhance cancer detection accuracy by developing a novel approach to feature extraction from cfDNA fragmentomics. MethodsWe implemented a supervised non-negative matrix factorization (SNMF) algorithm to generate embedding vectors capturing cancer-specific signals within cfDNA fragmentomic features. These embeddings served as input for a machine learning model to classify cancer patients from healthy individuals. ResultsWe validated our framework using two datasets: an in-house cohort of 431 cancer patients and 442 healthy individuals (dataset 1), and a published cohort comprising 90 hepatocellular carcinoma (HCC) patients and 103 individuals with cirrhosis or hepatitis B (dataset 2). In dataset 1, we achieved an AUC of 94% in pan-cancer detection. In dataset 2, our framework achieved an AUC of 100% for HCC vs healthy classification, 99% for HCC vs non-HCC patients classification, and 96% for identifying HCC patients among a mixed group of non-HCC patients and healthy donors. ConclusionThis study demonstrates the efficiency of SNMF-transformed features in improving both pan-cancer detection and specific HCC detection. Our approach offers a significant advancement in leveraging cfDNA fragmentomics for early cancer detection, potentially enhancing diagnostic accuracy in clinical settings.

bioinformatics↗

bioRxiv2023

Tissue of origin detection for cancer tumor using low-depth cfDNA samples through combination of tumor-specific methylation atlas and genome-wide methylation density in graph convolutional neural networks

BackgroundCell free DNA (cfDNA)-based assays hold great potential in detecting early cancer signals yet determining the tissue-of-origin (TOO) for cancer signals remains a challenging task. Here, we investigated the contribution of a methylation atlas to TOO detection in low depth cfDNA samples. MethodsWe constructed a tumor-specific methylation atlas (TSMA) using whole-genome bisulfite sequencing (WGBS) data from five types of tumor tissues (breast, colorectal, gastric, liver and lung cancer) and paired white blood cells (WBC). TSMA was used with a non-negative least square matrix factorization (NNLS) deconvolution algorithm to identify the abundance of tumor tissue types in a WGBS sample. We showed that TSMA worked well with tumor tissue but struggled with cfDNA samples due to the overwhelming amount of WBC-derived DNA. To construct a model for TOO, we adopted the multi-modal strategy and used as inputs the combination of deconvolution scores from TSMA with other features of cfDNA. ResultsOur final model comprised of a graph convolutional neural network using deconvolution scores and genome-wide methylation density features, which achieved an accuracy of 69% in a held-out validation dataset of 239 low-depth cfDNA samples. ConclusionsIn conclusion, we have demonstrated that our TSMA in combination with other cfDNA features can improve TOO detection in low-depth cfDNA samples.

cancer biology↗