An AI-based liquid biopsy test using genome-wide cell-free DNA (cfDNA) fragmentation patterns detected early liver fibrosis and cirrhosis, and may also reveal signals of broader chronic disease burden, according to research published in Science Translational Medicine.
To learn more about the fragmentome and its implications for diagnostics, we connected with lead researchers Victor Velculescu and Akshaya Annapragada at the Johns Hopkins Kimmel Cancer Center, USA.
Could you explain the concept of the fragmentome, and how it’s expanding our understanding of chronic diseases?
The cell-free DNA (cfDNA) fragmentome comprises tens of millions of DNA fragments that are released into the circulation from dying cells, both healthy and diseased, and can be assessed from less than 1 mL of blood using affordable, low coverage, whole genome sequencing. The genome-wide distribution, size and composition of these fragments reflect the underlying chromatin, genomic, epigenomic, and transcriptomic state of the cells they come from, allowing for an accessible, noninvasive “physiological snapshot”.
Many diseases have characteristic changes to the genome and epigenome in both organ-tissue derived cells and immune cells, and tracing these patterns in the blood can help identify and monitor disease, as well as generate biological insights.
How does AI help to interpret fragmentome data and generate actionable insights?
The human genome comprises 3 billion base pairs, a high-dimensional and large source of information, which AI is perfectly suited to analyze. Rather than looking for changes to individual genes, we use AI to make sense of nuanced changes across the entire genome. This gives us more shots on goal when detecting early-stage disease.
Moreover, AI allows us to analyze low coverage sequencing data which is far more affordable to obtain than typical deep coverage sequencing approaches. Our approach requires sequencing the whole genome a few times, while other approaches often sequence specific regions thousands to tens of thousands of times. This accessibility allows us to envision population-scale use of our approach in the US and global settings.
What did your study reveal about the potential of this technology to detect liver fibrosis from liquid biopsies?
Our approach identified more than half of individuals with early-stage liver disease – including steatosis, hepatitis, and early fibrosis – and more than three in four individuals with advanced fibrosis or cirrhosis. We saw limited false positive results and low cross-reactivity for other diseases.
This is especially exciting because existing blood biomarkers generally cannot pick up early liver disease and, even in advanced fibrosis and cirrhosis, our approach enabled improved detection.
This study is a proof-of-concept, and much work remains to advance it to the clinic, but it represents a significant milestone – one of the first demonstrations of genome-wide liquid biopsies for use in a noncancerous, chronic condition.
What advantages does your method offer over other liquid biopsy technologies?
Our approach truly analyzes the entire genome, including changes to fragment length, fragment distribution, epigenetic marks, and even the so-called “dark genome” of repeat elements. This contrasts with first-generation liquid biopsies that analyzed mutations or methylation changes in specific genes.
We believe this genome-wide approach allows us to capture a far larger set of potential alterations, opening the door to more sensitive, early detection. By using low coverage sequencing, we lower the expected cost of our assay – from $1000s for conventional liquid biopsies to $100s for our approach.
Finally, our AI-based approach harnesses thousands of features obtained from the same facile laboratory protocol. There’s a lot of headroom here, and this versatile platform allows for future innovation to identify additional genome-wide biomarkers.
What are the potential implications of this technology for patients with liver disease?
In the US alone, more than 100 million individuals are at risk for liver cirrhosis and cancer, yet many do not know it. Early-stage liver disease, or fibrosis, occurs prior to cirrhosis or cancer and is a good target for clinical intervention with lifestyle modifications and new drugs that can reverse fibrosis. However, current blood biomarkers rarely detect early fibrosis and miss cirrhosis about half the time.
Diagnostic imaging is available but has variable performance across communities and may be inaccessible. As metabolic conditions like obesity, hypertension, and diabetes continue to increase in prevalence, the number of individuals with undiagnosed liver disease will only increase. There is a pressing need to identify individuals with early disease and intervene rapidly – the exact scenario we envision for our technology.
What other conditions might be detectable with this technology?
We believe that our approach provides a comprehensive picture of the body in health and disease. In the present work, we provided a proof-of-concept for the use of this technology in liver disease, and also showed preliminary results in inflammatory, vascular, and neurodegenerative conditions, as well as for overall morbidity prediction.
Our group’s prior work has shown great potential for detection of several cancer types, including lung, liver, and ovarian tumors, and recently culminated in clinical validation and laboratory approval of the first liquid biopsy for lung cancer screening.
A key strength of our approach is its disease specificity – each of these models focuses on different sets of genomic features and is not cross-reactive for other diseases. Since essentially all cells in the body contribute cfDNA fragments at some level, the possibility exists to detect nearly any disease you can imagine.
A great deal of work is needed to thoroughly characterize the fragmentome in each disease process, and to determine the circumstances in which a disease-specific signal can truly be identified. We hope this study opens the door to future exploratory work and validation to realize the full potential of fragmentomic analyses across clinical applications.
Newsletters
Receive the latest pathologist news, personalities, education, and career development – weekly to your inbox.
