May 6, 2025 | Treatment for several solid tumors—notably colorectal, gastrointestinal, head and neck, and certain breast cancers—often includes a chemotherapy drug known as 5-fluorouracil (5-FU). It has been a standard and reliable component of drug cocktails for over 50 years, but it can be hazardous addition for the one in three patients harboring a genetic mutation putting them at risk of a severe and sometimes deadly reaction to the medicine, says Carla Finkielstein, Ph.D., professor at the Fralin Biomedical Research Institute at Virginia Tech Carilion.
Use of 5-FU “varies depending on the tumor type, molecular profile, and treatment guidelines,” she says. “But it remains a backbone agent in many standard chemotherapeutic regimens for solid malignancies.”
But it won’t work for patients with one of the 30 or so problematic mutations to the DPYD gene, which is responsible for providing instructions for the creation of DPD (dihydropyrimidine dehydrogenase), an enzyme crucial for breaking down 5-FU. Some of these mutations can cause serious side effects, or even death, if the body is unable to properly process the drug. The problem, according to Finkielstein, is that most oncologists, at least in the U.S., still do not routinely screen cancer patients for DPYD mutations prior to starting treatment.
That’s not the case elsewhere in the world, notably France and The Netherlands, where testing of the gene is required. In 5-FU package inserts, the U.S. Food and Drug Administration “recommends” such testing but without the teeth of a mandate uptake in clinical practice has been slow, says Finkielstein.
There is now plenty of evidence to conclude that individuals harboring certain DPYD variants are at increased risk of severe toxicity when administered standard doses of the chemotherapy drug (Journal of Clinical Oncology, DOI: 10.1200/JCO.22.02364). The testing is commonly done using a blood sample, but saliva testing would be a less invasive, patient-friendly alternative for patients already dealing with the emotional upheaval of a cancer diagnosis, Finkielstein says.
The analysis of saliva is more work for molecular biologists like herself because the fluid is “not as clean as blood” due to the presence of non-human DNA and bacterial contamination. “That is our problem to solve,” she says.
Finkielstein and her colleagues recently analyzed DNA from healthy volunteers and colon cancer patients to show that saliva samples could reliably detect known DPYD mutations. They also identified several new pathogenic variants after employing artificial intelligence (AI) tools, alongside 3D protein modeling, to evaluate thousands of other patient samples.
Preliminary findings were presented at the 2025 ASCO Gastrointestinal Cancers Symposium. An abstract also published in the Journal of Clinical Oncology (DOI: 10.1200/JCO.2025.43.4_suppl.226).
Genetic testing using saliva rather than blood is more economical because it comes with none of the costs associated with blood draws, including trained phlebotomists, medical supplies, shipping, and specialized storage, says Finkielstein. The cost of blood-based testing is one of multiple factors limiting widespread adoption of DPYD testing programs.
DPYD testing identifies the “molecular footprint” that can tell oncologists if the side effects of treatment with 5-FU are likely to outweigh the benefits, Finkielstein says. If none of the problematic mutations are discovered, patients can be given the drug knowing that their body will excrete most of the administered dose within 24 hours.
Not all genetic variations in the DPYD gene cause this problem, she notes, and some mutations are completely innocuous. But if a mutation leads to partial reduction or complete loss of DPD function, the body cannot effectively break down the chemotherapy agent in the liver. As a result, the drug builds up in the bloodstream, leading to elevated systemic exposure that can trigger severe side effects.
Despite its clinical importance, precautionary testing for DPYD mutations is still “not on the radar” of many healthcare professionals. Finkielstein estimates that no more than 5% of oncologists routinely order the test before initiating treatment.
“I became aware of the dilemma after sitting in on meetings where clinicians were presenting cases involving a DPYD mutation that, unfortunately, was only identified after treatment had already begun,” Finkielstein recalls. “That’s what motivated us to develop the best possible tool to support oncologists in doing what they care most about, helping their patients. Our goal was to bridge the gap between pathology and oncology to support both ends of the spectrum of care.”
“The test is not expensive, and this is not rocket science, but it can save lives,” she says. “If we have a tool that is simple to use and has the power to protect patients, why wouldn’t we use it? We owe them that.”
Any hospital or certified clinical laboratory equipped for genetic testing could process the samples. The relevant DPYD variants could simply be integrated into an existing pharmacogenetic testing panel.
Saliva samples might logically be taken at patients’ first visit to their oncologist when all options are still on the table, says Finkielstein. By the time they come back for their second visit, genetic test results would be back to guide treatment decision-making.
If genetic testing finds that a patient carries one mutated copy of the DPYD gene, evidence supports reducing the dose of the 5-FU rather than eliminating its use entirely, says Finkielstein. However, if both copies of the gene are mutated, it is “strongly recommended to avoid 5-FU altogether due to the high risk of severe or fatal toxicity.” Even with a dose reduction by up to 90%, patients with two deficient copies of the gene remain at significant risk, and clinical guidelines in countries like France and The Netherlands generally advice against using fluorpyrimidines (e.g., 5-FU) in these cases.
Although 5-FU is used in most chemotherapy drug cocktails for several solid tumors, alternative treatments are available for patients with predicted poor tolerance, she adds. But these options, which may include other chemotherapeutic drugs or targeted therapies, are often more expensive compared to the older workhorse drug.
Finkielstein and her team are using artificial intelligence to scan the genomes of over half a million people (some with cancer, some without) looking for DPYD mutations that might predict adverse reactions to 5-FU. By combining massive genetic databases with 3D models of the DPD protein, they can determine which mutations disrupt how the protein works, she says.
In automotive terms, Finkielstein likens the approach to finding a missing tire or brake pedal versus an irrelevancy such as a broken interior light. It offers “a powerful framework for turning vast amounts of genetic data into actionable clinical insights.”
DPYD is a “very large gene” relative to other genes in the human genome, meaning there are plenty of opportunities for mutations to occur across its sequence, says Finkielstein. There is no reason to assume all the troublesome variations have even been identified. “We may be unaware of additional deleterious mutations because, in cases where a patient dies as a result of treatment and no genetic testing was performed, that information is lost,” she adds, pointing to the promise of computational tools for predicting the functional impact of novel variants.
Research has focused on saliva samples from colon cancer patients, which is a significant health concern in both Virginia and West Virginia, but DPYD testing has broader, pan-cancer relevance, says Finkielstein. “The presence of [DPYD] mutations in patients, even when identified in the context of one cancer type, has implications for all cancers treated with fluorpyrimidine-based therapies.
“With the tools we have today to ensure safer therapies, this goal is no longer a dream,” she adds. “My aim is for every patient undergoing cancer treatment to remain a fully functional member of our society. As basic scientists, it's easy to get absorbed in technical work and lose sight of the human impact. But we have a responsibility to humanize our science and align it with the real needs of patients.”