Cell-Free DNA Fragmentomics for Liver Cancer Screening

Dr. Anna Pagoda from Johns Hopkins discusses using cell-free DNA fragmentomics to improve liver cancer screening.
Aims to detect tumors earlier and identify precancerous conditions like liver cirrhosis.

Unlike many tumor types, the incidence and mortality of liver cancer have been increasing.
This increase is attributed to a large global unscreened population.
Approximately 400 million people are at risk due to viral hepatitis or liver cirrhosis.
Less than 20% receive guideline-adherent screening (biannual ultrasound and alpha-fetoprotein).
A significant portion of the US population with metabolic-associated steatosis remains unscreened.

Early identification of liver cancer presents a significant public health opportunity.
Liver disease exists on a continuum, with early stages potentially reversible.
Individuals with cirrhosis and fibrosis have about a 30% lifetime risk of developing liver cancer.
Identifying individuals before decompensation is challenging.

Non-invasive cancer and precancer detection uses cell-free DNA fragmentomics.
Fragmentome: full compendium of DNA fragments of nucleosomal origin in the blood.
Reflects genomic, epigenomic, and chromatin alterations in tumors and early disease.
Signal extracted from low coverage whole genome sequencing.
Allows capturing multiple alterations in an accessible format.

Examines length and representation of nucleosomal DNA fragments across the genome.
Illustrates disorganization of chromatin and nucleosomal structure in cancer.

Focuses on the "dark genome" of repeat elements (lines, signs, satellites).
These elements implicated in cancer but often excluded due to technical challenges.
Measures about 1280 different types of repeat elements using an alignment-free, k-mer based approach.

Study on individuals with no liver disease, high-risk conditions (viral hepatitis, cirrhosis), and hepatocellular carcinoma (HCC).
Individuals without liver disease had consistent fragmentation profiles.
Liver cancer patients showed significant aberration.
Fragmentation profiles detected all stages of liver cancer with high sensitivity and specificity in both average and high-risk populations.
Performance was replicated in an external validation cohort.

Exploration of how early the signal can be detected.
Cirrhosis and viral hepatitis show fragmentomic signals.
Case-control cohort: no liver disease, pre-fibrotic liver disease, fibrosis, and cirrhosis.
Measured fragmentation profiles, repeat landscapes, and epigenetic profiles.
Cross-validated and locked model for cirrhosis detection, evaluated in an external validation cohort.

Enrichment of shorter, sub-mononucleosomal fragments in cirrhosis.
Healthy individuals show consistent fragmentation profiles, homogenous repeat landscapes, and epigenetic profiles.
Liver cirrhosis shows heterogeneity and disorganization in all feature classes.
Model cross-validated and locked for non-invasive detection of liver cirrhosis using Artemis and Delphi.

Model scores recapitulated the biology of liver disease.
Lowest scores in no liver disease, increased scores in pre-fibrotic disease, and highest scores in advanced fibrosis and cirrhosis.
Strong classification performance in both pre-fibrotic and fibrotic/cirrhotic settings.
Stair-step pattern observed in the external validation cohort.

Limited cross-reactivity observed in individuals with other fibrosis-related diseases (chronic pancreatitis and benign lung nodules).
Suggests liver disease-specific signal.

Fragmentomic test as a facile entry point to HCC screening.
Initial test for liver cancer detection in high-risk individuals, serving as a rule-in or prescreen to imaging and further diagnostic workup.
Fragmentomic test based on Artemis and Delphi for general populations, including those with steatosis and other high-risk conditions.
Rules in individuals for liver elastography, other imaging, and comprehensive disease management, including HCC surveillance.

Cell-free DNA fragmentomes are altered in individuals with liver and other cancers, as well as high-risk liver diseases.
Artemis and Delphi approaches can be adapted for early cancer and precancer detection.
Fragmentomic screening could enhance liver cirrhosis detection and target individuals for effective HCC screening.
Efforts underway to translate this approach in diverse clinical settings.

Thanks to the Cancer Genomics Lab at Hopkins, advisors Victor Valcolescu and Rob Scharf, Zach Foda, collaborators, the Romanian Genomics Institute, and the NIH.
Funding from NIH F30 fellowship and the family of Dan Zhang Scholar in Training Award.

Q: How the Kupffer cells responsible for clearing cell free DNA can affect the cell free DNA yield that you get from liver disease
A: All disease processes that alter cell free DNA is in some way they depend on liver function, because the liver is clearing cell free DNA

Q: Can you differentiate the viral hepatitis from steatoctatic fatty liver disease, from alcoholic cirrhosis
A: They can detect cirrhosis across several of these etiologies and it would be great to look further into how these fragmentation profiles might differ based on the underlying pathology.

Q: Quality Control Question for suboptimal specimen (too much lysis of the white blood cells)
A: Samples that do not have sufficient nucleosomal DNA are excluded and Variations of gel electrophoresis like a tape station is used to analyze what the size profile looks like prior to even going to sequencing

Q: if things change, you get new information over time when do you change the locked down algorithm?
A: machine learning approaches enable continuous learning and tuning of the weights, but we have not implemented one and we've generally stuck to the principle of locking a model in a discovery cohort and then validating it.

Q: Can you reverse the cirrhosis with the DNA levels?
A: There is scope to investigate that in cirrhosis, and certainly the extent of liver damage has some correlation to the scores that we were getting.