NEW YORK – A new artificial intelligence-driven clinical decision support tool aims to guide the choice of frontline therapy for pancreatic cancer patients and, its developers say, underscores the importance of next-generation sequencing for newly diagnosed patients.

Using its AI-driven precision medicine computational platform, Perthera has developed a model to predict patient response to each of the two standard frontline therapies for pancreatic ductal adenocarcinoma (PDAC): FOLFIRINOX and gemcitabine plus protein-bound paclitaxel. The choice of which regimen to try is currently made based on the patient’s history and the expected toxic effects of each treatment approach.

The predictor, dubbed PDACai, was developed within Perthera’s analytic engine, which uses a large language model to incorporate relevant scientific literature, clinical trial results, and individual patient information, including genetic testing results, routine labs and tests, and unstructured notes. Within the platform, machine-learning algorithms rank approved therapies and, in some cases, expanded use therapies, depending on the tumor and cancer type for an individual patient, and generate a list of personalized treatment options, dubbed the Perthera Report. Perthera staff manually check all patient data against the records to ensure it is captured correctly, and a virtual molecular tumor board reviews the treatment rankings, adjusting if necessary. Any adjustments are then added back into the model. Perthera is incorporating the PDACai into a new version of the report, Perthera PV (Pancreatic Version), which will be released in late February.

The McLean, Virginia-based firm developed the predictive PDACai signature using a proprietary repository of pancreatic cancer data developed with the Pancreatic Cancer Action Network’s Know Your Tumor program and in collaboration with researchers at other institutions conducting precision oncology studies. Last month, Perthera presented results from a validation study at the 2025 ASCO Gastrointestinal Cancers Symposium demonstrating that predictions made by PDACai correlated with overall survival and progression-free survival outcomes for first-line patients with metastatic PDAC on FOLFIRINOX or gemcitabine-paclitaxel. According to the researchers, with upfront next-generation sequencing, it is now feasible to use PDACai to guide first-line treatment decisions in this treatment setting.

Patients with pancreatic cancer have a five-year survival rate of 13 percent across all stages, and just 3 percent of those with distant metastases survive five years. In the US nearly 68,000 people are diagnosed with pancreatic cancer each year, and about 52,000 will die from it. For patients with advanced PDAC, the most common form of pancreatic cancer, the introduction of FOLFIRINOX and gemcitabine-paclitaxel as standard first-line therapies only improved median overall survival in that patient population by about three months between 1986 and 2016, to a median of about seven months.

Flavio Rocha, a surgical oncologist at Oregon Health and Science University (OHSU) and an investigator on the study, noted that a tool such as PDACai is important because nearly 50 percent of patients with metastatic pancreatic cancer die before they have an opportunity to receive a second-line therapy, and there are no head-to-head studies indicating which of the two standard options is superior.

Currently, oncologists make an educated guess as to which treatment to use based on clinical trials that have compared FOLFIRINOX to gemcitabine, and gemcitabine-paclitaxel to gemcitabine. “But you can’t do cross-trial comparisons. Different trials have different inclusion criteria, so you can’t really compare apples to oranges,” Rocha said. “Not only are we wasting time by not giving the optimal regimen, we’re also potentially exposing patients to toxicities that they may not need and delaying when they’ll they switch to the second-line therapy, which may be better for them.”

Researchers from Perthera and academic research centers including OHSU, NYU Langone Health, Cedars-Sinai Medical Center, and others are now working to validate PDACai predictions for second-line FOLFIRI versus FOLFOX chemotherapy. Beyond that, they hope to validate the model in prospective clinical studies.

Rocha envisions PDACai being used in the future to recommend biomarker-driven therapies as they are approved and become available for patients with pancreatic cancer. As an example, he noted that pancreatic cancer patients with hereditary cancer syndromes, such as BRCA1/2 mutations, are known to benefit from platinum-based chemotherapy and that there are ongoing clinical trials of PARP inhibitors for that patient population. “What’s really exciting is that 90 percent of pancreatic cancers have a KRAS mutation, [and] there’s now a whole slew of KRAS inhibitors coming online,” Rocha said. “It’s a very timely introduction of [AI] technology as the drugs come online, to see how we can use them more effectively.”

Rocha also speculated that PDACai could have utility to shrink tumors ahead of surgery for patients with localized disease. “If we can use this platform to try to predict which tumor will shrink more and get patients to surgery sooner, that would be ideal,” Rocha said. In that setting, time is of the essence. Rocha said only about half of patients with localized pancreatic cancer end up receiving postoperative chemotherapy, “which is even more reason to give [chemotherapy] beforehand and hopefully give the right one.”

In late February, Perthera will release Perthera PV, the newest version of its individualized AI-driven tool for optimizing treatment in solid tumors, the Perthera Report. Perthera PV, which will incorporate the PDACai, provides a list of ranked therapy options that are personalized to each patient, and, according to the company, will include recommendations beyond first-line therapy for pancreatic cancer, such as adjuvant, neoadjuvant, and later-line treatment suggestions.

Perthera President and CEO Donna Tuths said more than 1,500 oncologists at 600 cancer centers across the US have ordered the current version of the Perthera Report, and the company is already receiving requests for Perthera PV ahead of its release.

“We are already typing a patient today in advance of the launch,” Tuths said. “Time matters more here than in almost any other cancer.”

Next-generation sequencing and molecular testing are currently not commonly offered to first-line pancreatic cancer patients, since there are no approved molecularly targeted treatments in that patient setting. The standard of care is chemotherapy, and until now such tests would not have helped doctors choose between the two options.

But now, because genomic profiling results are part of the inputs for the AI analysis, Tuths argues the availability of the PDACai within Perthera PV offers “a compelling reason for pancreatic [cancer] patients to get [NGS] and molecular testing right upfront, because that can now be used immediately to guide their frontline treatment.”

Tuths noted that now that Perthera has validated the PDACai signature, the company aims to add similar specialized predictors in indications like lung, breast, and colorectal cancer to the pan-tumor Perthera Report.

View full article