Early Detection of Human Cancers Using a Cost-Effective, Non-Invasive Assay of the cfDNA Fragmentome

Friday, October 13, 2023

8:00 AM – 8:15 AM PDT

Location: 341 (Center, Level 3)

Presenting Author(s)

Akshaya Annapragada, AB;SM (she/her/hers)

MD/PhD Student
Johns Hopkins University
Baltimore, Maryland, United States

Co-Author(s)

JM

Jamie Medina

Johns Hopkins University School of Medicine, United States
Zachariah Foda, MD, PhD

Assistant professor
Johns Hopkins University School of Medicine, United States
PL

Pien Lof

The Netherlands Cancer Institute, United States
DM

Dimitrios Mathios

The Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, MD, USA, United States
SS

Sarah Short

Johns Hopkins University School of Medicine, United States
Adrianna Bartolomucci (they/them/theirs)

Research Specialist II
Johns Hopkins University School of Medicine
Baltimore, Maryland, United States
DB

Daniel Bruhm

The Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, MD, USA, United States
NV

Nicholas Vulpescu

The Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, MD, USA, United States
EJ

Euihye Jung

University of Pennsylvania Perelman School of Medicine, United States
VA

Vilmos Adleff

Johns Hopkins University School of Medicine, United States
HS

Heather Symecko

University of Pennsylvania Perelman School of Medicine, United States
HL

Harry Luu

The Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, MD, USA, United States
Dv

Daan van de broek

Department of Laboratory Medicine, Antoni van Leeuwenhoek Hospital – The Netherlands Cancer Institute, Amsterdam, The Netherlands, United States
SB

Stephen Baylin

The Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, MD, USA, United States
MP

Michael Press

Department of Pathology, Keck School of Medicine, University of Southern California, Los Angeles, CA, USA, United States
DS

Dennis Slamon

David Geffen School of Medicine, University of California Los Angeles, Los Angeles, CA, USA, United States
GK

Gottffried Konecny

David Geffen School of Medicine, University of California Los Angeles, Los Angeles, CA, USA, United States
SD

Susan Domchek

University of Pennsylvania Perelman School of Medicine, United States
RD

Ronny Drapkin

University of Pennsylvania Perelman School of Medicine, United States
AK

Amy Kim

Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States
JP

Jillian Phallen

Johns Hopkins University School of Medicine, United States
RS

Robert Scharpf

Johns Hopkins University School of Medicine, United States
CL

Christianne Lok

The Netherlands Cancer Institute, United States

Last Author(s)

VV

Victor Velculescu

Johns Hopkins University School of Medicine, United States

Introduction:: A significant amount of morbidity and mortality from cancer is attributable to late-stage diagnosis when treatments are less effective. While population screening or high-risk surveillance is recommended for some cancers, challenges to accessibility and effectiveness limit wide-spread use. Liquid biopsies of cell-free DNA (cfDNA) are a non-invasive option for earlier detection but are hampered by the limited number of tumor-specific changes detectable in the circulation. cfDNA fragments reflect the underlying chromatin structure of the genome in the cells they originate from, and alterations to their size and representation reflect cancer-related genetic and epigenetic changes. We previously developed DELFI (DNA Evaluation of Fragments for Early Interception) [1,2,3]. DELFI uses cost-effective, low-coverage, whole-genome sequencing and machine learning to evaluate millions of cfDNA fragments genome-wide obtained from as little as 1mL of blood. Here, we analyze cfDNA from 1,608 individuals to evaluate DELFI in three distinct clinical contexts: (a) Population-wide screening of ovarian cancer, a disease in which no clinical screening framework exists and only 20% of cancers are detected at an early stage (five-year survival 94% vs. 31% for late stage) [4]; (b) Surveillance of individuals at high-risk of liver cancer, where < 2% of patients receive guideline-adherent screening despite 11x higher survival for patients with local disease vs. distant metastases [2]; (c) Post-transplant monitoring of liver cancer patients treated by transplantation, a group where 15-20% experience recurrence and prognosis improves with timely identification [5]. Our results demonstrate the broad applicability of DELFI and its potential to improve early detection of cancer.

Materials and Methods:: We performed whole genome sequencing at 1-2X coverage on cfDNA extracted from the blood of 239 patients with ovarian cancer and 571 women at average risk of cancer in the US/EU and partitioned the cohort by institution to create an external validation cohort. We sequenced 75 patients with liver cancer, 133 patients at high-risk of liver cancer (78 cirrhotic, 55 viral hepatitis), and 293 men and women at average risk of cancer in the US/EU (including 146 women from the ovarian cohort) [2]. We obtained an external validation cohort (n=223) from a liver cancer study in Hong-Kong [6]. We computed the ratio of short (100-150 bp) to long (151-220 bp) cfDNA fragments in 473 non-overlapping 5Mb bins genome-wide, and quantified aneuploidy on each chromosomal arm as a z-score with respect to a panel of healthy individuals. We used these features to cross-validate and externally validate DELFI machine learning models (logistic regression and gradient boosted models) for ovarian and liver cancer screening and liver cancer surveillance. For 16 patients with liver cancer who received a liver transplant, we assessed post-transplant cfDNA samples. Finally, in our liver and ovarian cancer cohorts we assessed fragment coverage at binding sites of 797 transcription factors and compared these to an additional previously analyzed lung cancer cohort (n=62 cancer, n=158 non-cancer) [3]. We used these metrics to generate ranked lists of transcription factor binding sites with altered fragmentation profiles in cancer for use in a Gene Set Enrichment Analysis of the DisGeNET database of gene–disease associations.

Results, Conclusions, and Discussions:: Used as a population screening test, DELFI detected ovarian cancer with an AUC of 0.85 (95% CI 0.81-0.88) in the cross-validated cohort and an AUC of 0.82 (95% CI 0.72-0.92) in the external validation cohort. DELFI detected liver cancer with an AUC of 0.98 (95% CI 0.97-0.99) in the screening setting and an AUC of 0.90 (95% CI 0.86-0.94) in the surveillance of patients at high-risk of liver cancer. In the external validation cohort, the locked model detected liver cancer with an AUC of 0.97 (95% CI 0.95-0.99). When the locked liver cancer DELFI model was applied to cancer patients after they received a liver transplant with curative intent, DELFI scores fell a median of 30.6% from their pre-transplant maximum (IQR 103.5%). This suggests that DELFI may be useful for post-transplant monitoring of recurrence. In samples from patients with ovarian, liver and lung cancers, changes in cfDNA fragment coverage at transcription factor binding sites reflected tissue-specific origins. In a gene set enrichment analysis of these changes, there were statistically significant enrichments for gene-disease associations relevant to ovarian, liver and lung neoplasms for the ovarian, liver and lung cohorts respectively. These results demonstrate the utility of DELFI across a range of clinical contexts and patient populations. DELFI has the potential to improve clinical management of patients with cancer or at high-risk of cancer, and to enable accessible, non-invasive, population-wide cancer screening.

Acknowledgements (Optional): :

References (Optional): : [1] Cristiano SC, Leal A, Phallen J, et al. Nature. 2019;570:385-9.
[2] Foda ZH, Annapragada AV, Boyapati K, et al. Cancer Discovery. 2023;13(3):616-31.
[3] Mathios D, Johansen JS, Cristiano SC, et al. Nat Commun. 2021;12:5060.
[4] American Cancer Society. “Early Detection, Diagnosis, and Staging” and “Survival Rates for Ovarian Cancer”. Accessed Apr 6, 2023.
[5] Filgueira NA. World J Hepatol. 2019;11(3):261-272.
[6] Jiang P, Chan CW, Chan KC, et al. PNAS. 2015;112:E1317–25.