Science & Enterprise logo
Science for business people. Enterprise for scientists.

A.I. Shown to Enhance Diagnostics in Liquid Biopsies

Digital DNA
(Pete Linforth, Pixabay. https://pixabay.com/illustrations/dna-life-biotechnology-evolution-4068826/)

9 Mar. 2023. A new analysis shows machine learning algorithms can provide more diagnostic details in liquid biopsies, blood tests to detect and track solid tumor cancers. Researchers from the biotechnology company Reveal Genomics S.L. in Barcelona, Spain and academic labs in Spain and the U.S. report their findings in the 1 March issue of the journal Nature Communications.

Liquid biopsies offer a less invasive technique for cancer diagnostics and guiding precision medicine treatments by analyzing blood samples of patients for genomic properties of their tumors rather than taking surgical tissue samples. Data from liquid biopsies can provide important details about proposed treatments, for example the likelihood of patients to respond to immunotherapies. Up to now, say the authors, circulating tumor DNA or ctDNA — pieces of tumors released into the bloodstream — do not offer as rich a source of genomic detail as tissue biopsies. Plus, in some cases of advanced-stage cancers, tissue biopsies may not be feasible.

Reveal Genomics is a three year-old enterprise developing cancer diagnostics with machine learning to boost the data available from conventional techniques, including tissue biopsy samples. The company focuses so far on breast cancer and offers a test to analyze patients expressing human epidermal growth factor receptor 2 or HER2-positive proteins found in 14 percent of breast cancer patients, according to National Cancer Institute. Reveal Genomics cites findings showing its HER2DX test returns results on four different gene signatures in HER2-positive that combined with clinical data can calculate risk of recurrence and likelihood of a complete response.

The company says it trains its HER2DX algorithms with data from publicly available genomic and cancer databases, as well as breast cancer clinical data sets and survival studies. Reveal Genomics says the algorithms are also validated in more than 2,000 cases.

Detailed DNA signatures similar to tissue biopsies

In the new paper, researchers sought to extend the company’s analytics to liquid biopsies from tissue samples, with the goal of extracting more detailed data about breast tumor characteristics for clinical decisions.  The team took blood plasma samples from 459 patients with metastatic or spreading breast cancer in Barcelona and Madrid. The patients represented different breast cancer types, but included 245 patients with hormone receptor-positive (HR+)/HER2-negative breast cancer taking endocrine therapy or receiving treatments that block cyclin-dependent kinase 4 and 6, or CDK4/6, enzymes promoting cell proliferation, often used with endocrine therapy. And the team collected breast tissue samples from 185 individuals, including 110 patients also providing blood samples.

The researchers conducted whole-genome sequencing of the blood plasma samples and applied 150 previously defined multi-factor DNA signatures for solid tumor cancers. The authors report algorithm-driven analytics applied to the blood plasma samples return detailed DNA signatures similar to results from tissue biopsies, including measures of tumor proliferation and estrogen receptor signals, often the target of endocrine therapies.

The results also reveal four more DNA-based breast cancer subtypes from the group of 150 DNA signatures applied to whole genome sequencing, associated with survival outcomes. In addition, the analysis highlights a particular genomic signature called retinoblastoma loss-of-heterozygosity or RB-LOH, highly associated with poor response and survival outcomes following certain breast cancer treatments.

Aleix Prat, the paper’s lead author, and Reveal Genomics co-founder and chief scientist, says in a company statement that the study shows the company’s technology for analyzing plasma samples is feasible. “Our proprietary and novel supervised learning computational approach,” notes Plat, “predicts complex tumor features including gene expression, protein and tumor histology using one data source: DNA sequencing.” Patricia Villagrasa, co-founder and CEO, says the company plans to add liquid-biopsy analytics as a new service. “This new ctDNA-based assay, called DNADX,” adds Villagrasa, “is our second product expected to be available in 2024.”

More from Science & Enterprise:

We designed Science & Enterprise for busy readers including investors, researchers, entrepreneurs, and students. Except for a narrow cookies and privacy strip for first-time visitors, we have no pop-ups blocking the entire page, nor distracting animated GIF graphics. If you want to subscribe for daily email alerts, you can do that here, or find the link in the upper left-hand corner of the desktop page. The site is free, with no paywall. But, of course, donations are gratefully accepted.

[wpedon id=”42724″ align=”center”]

*     *     *