An innovative artificial intelligence (AI) technique for identifying biomarkers in tumor biopsies offers the potential to significantly reduce the time and costs associated with cancer diagnostics, thereby making precision oncology more accessible for patients who are underserved or lack resources.
Researchers from the University of California San Diego, comprising engineers and medical experts, have created a new generation of AI tools that enable rapid and cost-effective detection of clinically significant genomic changes directly from tumor biopsy slides.
A recent publication in the Journal of Clinical Oncology details this novel AI protocol, known as DeepHRD, aimed at streamlining the process of analyzing routine biopsies.
According to senior author Ludmil Alexandrov, Ph.D., who is a professor of bioengineering and cellular and molecular medicine at UC San Diego, this approach is primarily focused on saving both time and money in the clinical oncology treatment processes for breast and ovarian cancers.
The research team asserts that their advancements mark a significant progress towards addressing the issues of treatment delays and health disparities that have hindered the effectiveness of precision medicine for cancer patients. In essence, they are working on creating new AI tools that could either supplement or replace the costly and protracted genomic testing essential for tailoring effective cancer treatments to individual patients.
“Cancer patients today typically endure significant wait times—often crucial weeks—after a tumor is diagnosed before undergoing a standard genomic test, which can lead to life-threatening delays in care,” said Alexandrov. “It is particularly alarming that the high costs and lengthy timelines make life-saving treatments out of reach for many, especially in resource-limited settings.”
This project at UC San Diego reflects a multidisciplinary collaboration across the campus, involving the Department of Cellular and Molecular Medicine, the Shu Chien-Gene Lay Department of Bioengineering, Institute of Engineering in Medicine, Department of Medicine, and the UC San Diego Moores Cancer Center.
The promise of precision oncology—customizing treatment options for individual patients—was a key motivator for the team, with Erik Bergstrom, Ph.D., the study’s lead author and postdoctoral researcher in Alexandrov’s lab, emphasizing the objective.
“Sadly, high expenses, tissue requirements, and slow turnaround times have limited the broader application of precision oncology, resulting in less than optimal and even harmful treatments for cancer patients,” stated Bergstrom. “Our goal was to devise a distinct method to tackle this important issue by designing AI that can bypass the need for genomic testing.”
Bergstrom noted that the team concentrated on making the most of the limited patient information typically available at the early stages of diagnosis. He explained that almost every cancer patient undergoes a tumor biopsy—a tissue sample that is routinely examined using a light microscope, a method established in the late 19th century which still serves as the foundational approach in clinical oncology
“Our AI can be applied directly to a standard tissue slide for quick and precise identification of cancer genomic biomarkers,” Bergstrom added. The team specifically focused on an AI tool that identifies biomarkers indicative of homologous recombination deficiency (HRD), a condition where cancer cells lose a critical DNA repair mechanism.
Bergstrom emphasized that patients with ovarian or breast cancers displaying HRD typically respond favorably to common chemotherapy treatments, such as platinum and PARP (poly-ADP ribose polymerase) therapies.
“This AI method significantly reduces waiting time for patients,” Alexandrov remarked. “Oncologists can initiate treatment immediately upon the initial tissue diagnosis. Notably, our AI test has an almost non-existent failure rate, whereas conventional genomic tests fail 20 to 30 percent of the time, leading to the necessity for re-testing or even invasive re-biopsies.”
Co-senior author Scott Lippman, M.D., a distinguished professor of medicine at UC San Diego, highlighted that the new technology would eliminate time and financial barriers, ensuring immediate and equitable access to actionable genomic biomarker detection, which is vital for precision therapy, among patients with advanced cancers. A remarkable aspect of this breakthrough is that it has the potential to benefit well-resourced populations while also addressing significant disparities in precision medicine, particularly in remote, resource-limited areas globally where testing is currently lacking.
“The era of precision oncology gained traction in the late 1990s, yet recent studies in the U.S. indicate that a large number of cancer patients are not receiving FDA-approved precision therapy,” Lippman stated. “The main reason for this is the lack of testing. As a clinical oncologist with nearly 40 years of experience, I firmly believe that this approach represents the future of precision oncology.”
The AI technology behind DeepHRD is currently secured by provisional patents held by UC San Diego, which have been licensed to io9, a firm significantly involved with Alexandrov, Bergstrom, and Lippman, aiming to expedite the clinical deployment of this AI platform. This initiative seeks to deliver precision therapy to cancer patients more swiftly by getting them on the right treatments sooner. The authors anticipate that this same technology could also be applied to various other genomic biomarkers and different cancer types.
Affiliations: The authors are associated with several departments and institutes, as mentioned: Ludmil B. Alexandrov is linked with UC San Diego Moores Cancer Center, School of Medicine, and the Department of Bioengineering, among others. Erik N. Bergstrom is affiliated with Moores Cancer Center and the Department of Cellular and Molecular Medicine, while Scott M. Lippman has connections with Moores Cancer Center and the Center for Engineering and Cancer. Additional co-authors include Ammal Abbasi and Marcos DÃaz-Gay from Moores Cancer Center and the Department of Cellular and Molecular Medicine; Loïck Galland and Sylvain Ladoire from the Department of Medical Oncology and the Platform of Transfer in Biological Oncology Centre, Georges-François Leclerc Cancer Center, and the University of Burgundy-Franche, France.
Funding: This research received funding from the National Institutes of Health through various grants, including R01ES032547 and U01DE033345 to Ludmil B. Alexandrov, as well as P30 CA023100 to Scott M. Lippman, in addition to support from a Curebound Targeted grant and start-up funding from UC San Diego to Alexandrov. The study was also backed by the UC San Diego Sanford Stem Cell Institute.
