TORONTO, January 04, 2022 – Early detection of primary liver cancer (PLC), including hepatocellular carcinoma (HCC), intrahepatic cholangiocarcinoma (ICC), and combined HCC-ICC (cHCC-ICC), is essential for optimal survival of patients. Up-to-date, a fast, affordable, and accurate model is still needed for PLC early detection. Geneseeq and Shanghai Fudan University, Zhongshan Hospital published a prospective study in Hepatology further expanding the usage of Geneseeq’s MERCURY technology, a novel multi-dimensional fragmentomics approach, in the DECIPHER (Detecting Early Cancer by Inspecting ctDNA Features)-liver program.
In this study, a total of 362 participants, including 192 PLC patients, 53 liver cirrhosis (LC) or hepatitis B virus (HBV) patients, and 117 healthy volunteers, were enrolled as the training cohort, which was used to construct a machine learning model. Plasma samples were collected from the participants followed by cell-free DNA (cfDNA) extraction and low-coverage whole-genome sequencing. Two cfDNA fragmentomics features, including fragment size ratio (FSR), fragment size distribution (FSD), were then extracted and employed by the ensemble stacked model incorporating three machine learning algorithms (gradient boosting machine, random forest, and deep learning). The model performance was then assessed in an independent testing cohort of 354 participants (189 PLC patients consisted of 157 HCC, 26 ICC, 6 cHCC-ICC, and 165 non-cancer controls).
The model showed excellent performance for cancer detection in the testing cohort (Area Under the Curve [AUC]:0.995, 96.8% sensitivity at 98.8% specificity). It also showed excellent sensitivities in detecting early-stages PLC (I: 95.9%, II: 97.9%), small tumors (<=3cm: 98.2%), and HCC (96.2%) or ICC (100%). The AUC for distinguishing PLC from LC/HBV reached 0.985 (96.8% specificity at 96.1% specificity). Promisingly, our model maintained consistent performances during the downsampling process, even using 1X coverage data (AUC: 0.994, 93.7% sensitivity at 98.8% specificity). A separate model for distinguishing ICC from HCC was also constructed using the same fragmentomic features, yielding an AUC of 0.776.
“An ultra-sensitive yet affordable non-invasive liquid-biopsy based assay can be used to massively screen high-risk population, which would greatly facilitate the diagnosis of HCC or PLC at an early stage”, says Dr. Bao Hua, author and the Director of Geneseeq‘s R&D Department. The research team from Geneseeq investigated various liquid biopsy-based early cancer screening technologies such as cfDNA mutation, methylation, fragmentomics characteristics, metagenomics, and developed the multi-omics MERCURY technology. Through collaborations with clinicians, Geneseeq has carried out a series of cancer early screening studies DECIPHER using MERCURY in common solid tumor types. The DECIPHER-colon study has also been recently published in the Journal of Hematology & Oncology, demonstrating excellent ability to detect early-stage colorectal cancer (Stage 0/I) and advanced colorectal adenoma.