NEW YORK – A Johns Hopkins University-led team has developed a circulating cell-free DNA (cfDNA) "fragmentome" and machine learning approach to boost hepatocellular carcinoma detection, suggesting that the strategy may help in monitoring individuals with viral hepatitis, cirrhosis, non-alcoholic fatty liver disease (NAFLD), and other risky conditions.
"Medical societies throughout the world recommend screening for the highest risk populations, currently with abdominal ultrasound imaging with or without alpha-fetoprotein," co-senior and co-corresponding authors Victor Velculescu, an oncology researcher and co-director of Johns Hopkins Kimmel Cancer Center's Cancer Genetics and Epigenetics program, and Johns Hopkins University School of Medicine's Amy Kim, and their colleagues explained in Cancer Discovery on Friday.
Even so, the authors noted that overall adherence to international guidelines remains low and emphasized the "great need for development of accessible and sensitive screening approaches for HCC worldwide."
For their latest study, the researchers relied on low-coverage whole-genome sequencing to profile cfDNA fragment features in blood samples from individuals with HCC, healthy control individuals, or individuals with other liver conditions such as viral hepatitis, liver cirrhosis, or NAFLD.
From there, they turned to an artificial intelligence approach known as DELFI (DNA evaluation of fragments for early interception) to focus in on the cfDNA fragmentome features corresponding to HCC — an approach that members of the same team used to come up with a lung cancer classifier in the past.
"Given the direct connection between genomic and chromatin changes in liver cancer and cfDNA fragmentation, we used a machine learning approach to determine if changes in cfDNA fragmentomes could distinguish patients with HCC from those without cancer," the authors explained.
The team began by applying the approach to blood plasma samples from more than 500 individuals treated in the US or Europe, bringing in available chromatin immunoprecipitation sequencing data on liver cancer transcription factor binding patterns to flag DNA fragmentome features that were distinct in plasma samples from 75 HCC patients compared to those from 133 high-risk participants with other liver conditions or from healthy control individuals.
In the process, the investigators identified genome, chromatin, and transcription factor binding site shifts that marked the DNA fragmentomes found in non-cancer plasma samples. While HCC-related fragmentomes were more diverse, they noted, it was possible to put together a classifier score with DELFI to distinguish HCC cases — results were subsequently validated with blood plasma samples from 223 more individuals from Hong Kong, including individuals with or without HCC.
In particular, the team explained, DELFI scores were significantly higher for individuals with HCC, regardless of tumor stage, while intermediate scores tended to turn up in individuals with hepatitis or cirrhosis. Individuals without HCC, hepatitis, or cirrhosis had low DELFI scores, on the other hand.
The researchers noted that the DELFI score could detect HCC with 88 percent sensitivity and 98 specificity in a group of individuals who appeared to be at average risk of HCC, while the sensitivity was 85 percent (with some 80 percent specificity) in the high-risk group.
"Increased early detection of liver cancer could save lives, but currently available screening tests are underutilized and miss many cancers," Velculescu said in a statement.
Velculescu is a founder, board member, and investor in a related spinout company known as Defli Diagnostics.
He and his colleagues cautioned that the current findings will need to be further validated in additional large studies. Still, the suggested, their observations that scalable, cost-effective, and non-invasive cfDNA fragmentome analyses can detect liver cancer patients "may provide an opportunity to screen high-risk and general populations worldwide."