Using Cell-free DNA as a Marker for Liver Cancer Diagnosis

During this year’s AACR annual meeting we spoke to Kang Zhang, M.D., Ph.D., Founding Director of the Genomic Institute at the University of California, San Diego about the collaborative work he has been doing with the Laboratory for Advanced Medicine (LAM). Their recent study findings on the use of a blood-based method to look at cell-free DNA (cfDNA) or circulating tumor DNA (ctDNA) as a marker for liver cancer diagnosis and prognosis were presented during the conference. During the interview Zhang highlights the incidence of liver cancer, he discusses how identifying methylation patterns is a superior approach to identifying mutation when it comes to cancer testing and expands on the role AI and deep learning played in the study.

LM: Could you tell us more about your professional background, the collaborative work you have been doing with the Laboratory for Advanced Medicine (LAM), and more specifically, the talk you presented at this year’s AACR conference?

Kang Zhang (KZ): I'm a geneticist, and also a physician scientist. I've been working with genomics and epigenomics for the past ten years and I have been working with LAM since last year.

Our AACR presentation discussed the use of blood plasma as a way of performing liquid biopsy, specifically use of a blood-based method to look at cell-free DNA (cfDNA) or circulating tumor DNA (ctDNA) as a marker for the diagnosis and prognosis of hepatocellular carcinoma (HCC) – liver cancer. This was done in collaboration with Sun Yat-sen University Cancer Center, which is one of the largest cancer hospitals in China.

Liver cancer is classed as one of the deadliest cancers and there are currently limited targeted therapies available to combat the disease. The incidence of cancer in China is particularly high – this is probably due to the huge population of high-risk patients being carriers of hepatitis as well as the high prevalence of liver cirrhosis.

Annually, there are approximately 782,000 people newly diagnosed with liver cancer worldwide, half of which are in China, and approximately 746,000 people die of liver cancer each year.¹ 

LM: When it comes to cancer testing, why is the identification of methylation patterns a superior approach to identifying mutations?

KZ: There are many ways that a mutation can cause something like a cancer state. For example, there are multiple – actually dozens – of driver mutations that can cause liver cancer. However, for liver cancer to occur, the cancer cell must manually methylate a precise DNA site.

When it comes to a genetic mutation; it might only occur in 20–30% of cancer cells in one tumor sample, or among different tumors of different patients, however certain methylations have to happen in all of the cancer cells – this essentially gives you a binary switch.

LM: LAM recently completed the world’s largest clinical trial with 23 common cancers and 20,000 patients, could tell us more about this study?

KZ: So far, this is the largest sample size used in a cell-free DNA based cancer study. We're also using a new technology called DNA methylation which is a very common modification of DNA. This is significant because methylation is a key epigenetic mechanism used by cells to control gene expression – allowing genes to be turned on and off by modifying the DNA strands.

For a normal cell to become a cancer cell, and then to maintain its cancer cell state, it must turn on or off some of their genes. An oncogene can be turned on by demethylating the promoter of the oncogene that altered DNA methylation patterns may be a good indicator of an emerging tumor. Also, hyper-methylation of tumor suppressor genes occurs in the early stages of tumor development, observing more methylation marks could therefore indicate the presence of an emerging tumor. So obviously, hunting down those markers will help to track the cancer status, meaning a staging as well.

LM: Could you expand on the artificial intelligence (AI) and deep learning work that was involved?

KZ: We used machine learning to extract a huge amount of data from the methylation patterns. If you look at the cells in the blood, most of them are coming from ‘normal’ cells in the background – from the white blood cells or red blood cells.

The question is: How would you specifically find tumor-derived DNA methylation markers, or a methylation pattern amongst the sea of the other normal cells? That really is a very big challenge – it’s sort of like finding a needle in a haystack.

The idea is that by using a combination of machine learning algorithms and a large sample size, you can tease out the true signal from the background ‘noise’ to identify the methylation markers that correlate to the presence of HCC.

LM: Is there anything you can tell me about next steps, or related projects with LAM that may be impending?

KZ: In addition to the liver cancer work I presented at AACR, the IvyGene test has been validated for other common cancers such as lung cancer, colorectal cancer and breast cancer and is now commercially available. The test’s soft launch sold more than 2,000 test kits and the number of test kits administered is growing quickly. 

Kang Zhang was speaking to Laura Elizabeth Mason, Science Writer for Technology Networks.