Lesion Shedding Model: unraveling site-specific contributions to ctDNA

Kahn Rhrissorrakrai, Filippo Utro, Chaya Levovitz, Laxmi ParidabioRxiv 2021 · DOI

Abstract

Sampling circulating tumor DNA (ctDNA) using liquid biopsies offers clinically important benefits for monitoring cancer progression. A single ctDNA sample represents a mixture of shed tumor DNA from all known and unknown lesions within a patient. Although shedding levels have been suggested to hold the key to identifying targetable lesions and uncovering treatment resistance mechanisms, the amount of DNA shed by any one specific lesion is still not well characterized. We designed the Lesion Shedding Model (LSM) to order lesions from the strongest to the poorest shedding for a given patient. By characterizing the lesion-specific ctDNA shedding levels, we can better understand the mechanisms of shedding and more accurately interpret ctDNA assays to improve their clinical impact. We verified the accuracy of the LSM under controlled conditions using a simulation approach as well as testing the model on three cancer patients. The LSM obtained an accurate partial order of the lesions according to their assigned shedding levels in simulations and its accuracy in identifying the top shedding lesion was not impacted by number of lesions. Applying LSM to three cancer patients, we found that indeed there were lesions that consistently shed more than others into the patients’ blood. In two of the patients, the top shedding lesion was one of the only clinically progressing lesions at the time of biopsy suggesting a connection between high ctDNA shedding and clinical progression. The LSM provides a much needed framework with which to understand ctDNA shedding and interpret ctDNA assays.

Introduction

Liquid biopsies, which sample circulating tumor DNA (ctDNA) from the blood, represent a significant advancement in non-invasive cancer monitoring. However, a major challenge remains: a blood sample is a "genomic soup" containing DNA from all lesions in a patient's body. While clinicians can identify specific mutations, determining their origin—whether from a stable nodule or a rapidly growing metastasis—is difficult. Because tumors shed DNA at varying rates due to factors like size, necrosis, or metabolic activity, the signal in the blood is often biased. The Lesion Shedding Model (LSM) addresses this by quantifying site-specific contributions, making liquid biopsies more actionable for identifying targetable lesions and treatment resistance.

Methods

The LSM uses whole-exome or genome sequencing data from multiple tissue biopsies and at least one blood sample to estimate relative shedding levels. The workflow involves several key steps:

  1. Hypothesis Blood (HB) Generation: The model creates thousands of synthetic blood profiles by assigning different shedding weights ($\alpha$) to known lesions using a Dirichlet process.
  2. Missing Lesion Modeling: To account for unbiopsied tumors, the model uses subsampling of available lesions.
  3. Optimization: The model compares the "Hypothesis Blood" to the actual patient blood using cumulative precision curves, minimizing the $L_\infty$ (Chebyshev) distance to find the best fit.
  4. Consensus Shedding Networks: Results from multiple subsampling runs are aggregated into a directed graph. An edge from Lesion A to Lesion B indicates that A consistently sheds more than B.
Workflow of the Lesion Shedding Model (LSM) for estimating site-specific DNA shedding.

Results

The study validated the LSM using both simulated and real patient data from metastatic gastrointestinal cancer cases (TPS037 and TPS177).

Simulation Accuracy and Metric Selection

In simulations, the LSM demonstrated high accuracy in recovering pre-set shedding levels. For patient TPS037 (4 lesions), the model achieved a Pearson correlation of 0.97 between assigned weights and predicted network edges. For the more complex TPS177 case (17 lesions), the correlation remained significant at 0.54. The choice of distance metric was critical; the $L_\infty$ (Chebyshev) distance provided the highest weighted mean correlation compared to other metrics like Euclidean or L1 distance.

Clinical Correlation and Lack of Genetic Bias

A key finding was that high shedding levels correlated more strongly with clinical progression than with tumor size or genetic uniqueness. In both patients, the "top shedder" identified by the LSM was the specific lesion clinically progressing at the time of biopsy. Furthermore, the model showed no correlation between a lesion's genomic distance (uniqueness) and its predicted shedding level, confirming that the LSM captures true shedding dynamics rather than genetic distinctness.

Validation of LSM performance across distance metrics and correlation between assigned weights and predicted shedding for patients TPS037 and TPS177.

Discussion

The LSM provides a robust framework for "unmixing" the complex signals in liquid biopsies. By identifying top-shedding lesions, it offers a clearer lens for interpreting ctDNA assays, particularly in identifying sites of treatment resistance. While the model currently requires invasive tissue biopsies, future iterations could potentially use Circulating Tumor Cells (CTCs) as a proxy. Larger cohort studies are needed to further validate the link between shedding and clinical progression across diverse cancer types.

Availability: The LSM binary is available at https://github.com/ComputationalGenomics/LSM.

Summary updated

Abstract

Abstract Sampling circulating tumor DNA (ctDNA) using liquid biopsies offers clinically important benefits for monitoring cancer progression. A single ctDNA sample represents a mixture of shed tumor DNA from all known and unknown lesions within a patient. Although shedding levels have been suggested to hold the key to identifying targetable lesions and uncovering treatment resistance mechanisms, the amount of DNA shed by any one specific lesion is still not well characterized. We designed the Lesion Shedding Model (LSM) to order lesions from the strongest to the poorest shedding for a given patient. By characterizing the lesion-specific ctDNA shedding levels, we can better understand the mechanisms of shedding and more accurately interpret ctDNA assays to improve their clinical impact. We verified the accuracy of the LSM under controlled conditions using a simulation approach as well as testing the model on three cancer patients. The LSM obtained an accurate partial order of the lesions according to their assigned shedding levels in simulations and its accuracy in identifying the top shedding lesion was not impacted by number of lesions. Applying LSM to three cancer patients, we found that indeed there were lesions that consistently shed more than others into the patients’ blood. In two of the patients, the top shedding lesion was one of the only clinically progressing lesions at the time of biopsy suggesting a connection between high ctDNA shedding and clinical progression. The LSM provides a much needed framework with which to understand ctDNA shedding and interpret ctDNA assays. Availability Binary is available at https://github.com/ComputationalGenomics/LSM
BibTeX
@misc{rhrissorrakrai2021lesion,
  author = {Kahn Rhrissorrakrai and Filippo Utro and Chaya Levovitz and Laxmi Parida},
  title = {{Lesion Shedding Model: unraveling site-specific contributions to ctDNA}},
  year = {2021},
  doi = {10.1101/2021.01.28.428297},
  howpublished = {bioRxiv},
}
Figure