Scientific illustration of a migrating breast cancer cell - 3d illustration
Credit: Christoph Burgstedt/Getty Images

GC Genome Corporation published this week in Nature Communications on the company’s novel AI-based liquid biopsy technology. The study highlights the technology’s accuracy in early cancer detection and tissue-of-origin localization. The approach uses advanced artificial intelligence (AI) algorithms to analyze mutation density and patterns of cell-free DNA (cfDNA) and epigenomes.

The study was carried out in collaboration with the Korea Advanced Institute of Science and Technology (KAIST). The lead author is Mingyun Bae, Department of Bio and Brain Engineering, KAST.

Cancer is estimated to be the first or second cause of death before age 70 in most countries. Early detection of various types of cancer is critical since this disease has a better prognosis and survival rate the sooner it is diagnosed and treated. Early detection remains challenging though.

Using cfDNA-based noninvasive cancer screening for Multi-Cancer Early Detection (MCED) and localization is one option. Dying cells release (cfDNA) into the bloodstream, which includes a small amount of circulating tumor DNA (ctDNA) that comes from tumor cells when they die.

But there are challenges. As the authors note, “Circulating tumor DNA (ctDNA) reflects tumor-specific genetic and epigenetic alterations. Also, ctDNA fragments are physically shorter than normal cfDNA fragments.”

They list several approaches that have been used to detect cancer ctDNA including: targeted approaches relying on deep sequencing of recurrent mutations, genome-wide methods looking at DNA methylation patterns,, and genomic fragmentation patterns coupled with copy number variations (CNVs) or with chromatin signatures.

The  study includes a total of 2,543 cancer patient samples and 1,241 normal control samples and describes an ensemble algorithm that incorporates genomic and epigenomic profiles of tumor tissues into a deep learning model. This model analyzes mutation distribution and chromatin organization in reference tumor tissues and uses them as model features to detect the existence of cancer and determine the type of cancer in cfDNA samples.

The researchers say the technology has demonstrated “promising results” in detecting multiple types of cancer at an early stage. It has shown exceptional sensitivity, achieving a 91.1% performance rate based on a 95% specificity and a high level of accuracy, with an 81.7% success rate in predicting both the presence and type of cancer.

“These results suggest that the GC Genome AI-based cancer early detection technology could help reduce cancer deaths by offering a convenient, high-performing test to people,” said Eun-Hae Cho, MD, Chief Technical Officer at GC Genome. “We will continue our research to improve accuracy and sensitivity and look forward to making our technology accessible to all cancer patients worldwide.”

“Since cancer screening technologies are limited or premature, diagnostic tests are usually performed after symptoms arise, and early intervention opportunities are often missed, leaving few treatment options,” said Jung Kyoon Choi, senior author of the paper, from the Department of Bio and Brain Engineering, KAIST.

“We hope that our methods leveraging large-scale tumor genomes and epigenomes as reference data lay the groundwork for accurate cfDNA-based cancer diagnosis at early stages,” Choi added.

GC Genome offers genetic diagnosis services for Oncology, Pre&Neonatal, Rare Diseases, and Health Check-ups. The company was established in 2013 as a subsidiary of the GC group.

Also of Interest