(UroToday.com) The 2026 GU ASCO annual meeting featured a Keynote Address by Dr. Eliezer Van Allen discussing reimaging cancer research and care in the age of artificial intelligence. Dr. Van Allen started his lecture by highlighting three key cornerstones to assessing treatments and optimal outcomes for patients and addressing how artificial intelligence may advance these goals:
- Why? Biological discovery
- For Whom? Clinical care
- How? Ethics and implementation
Starting with the Why?, an example is learning biology for cancer discovery. We may start with >1,000 prostate cancer genomes, but there is an unknown black box in the middle that leads to who has primary versus lethal forms of prostate cancer:
However, once we look into the black box, there may be genes, pathways, and biological pathways whereby “biologically-informed” artificial intelligence is fully interpretable to enable target discovery and biomarker development, which may lead to new trial opportunities:1
There are several ways to do this, including (i) single cell foundation models, (ii) better deep learning architectures, and (iii) new biological priors via large language models and agents, the latter of which Dr. Van Allen further expanded on. At a basic level, large language models take in input (text), followed by learning patterns from data, subsequently providing an output (text):
Dr. Van Allen notes that when moving from large language models to artificial intelligence agents for asking Why?, the large language model needs to take action (input to tools), interact with the environment, and provide feedback (output from tools), while also providing a level of reasoning. Notably, human input reasoning is not absolutely necessary:
How we test this new artificial intelligence approach for representing all cancer biology is at the core of the work that Dr. Van Allen and his lab are currently working on.
Next, Dr. Allen focused on For Whom?, which has to date focused on using artificial intelligence to develop novel biomarkers. For example, this may include pathology, radiology, and/or molecular biomarkers:
However, Dr. Van Allen notes that these biomarkers are currently specialty specific and static, and we need to move to more general and active biomarkers with artificial intelligence. In work from Dr. Van Allen’s group,2 the reported that for evaluating clinical trends in groups of similar patients, genomic embeddings from all patients can be clustered, and clusters are interpreted based on their defining genomic biomarkers, as well as whether they separate patients by response to specific therapy types:
Next, the nearest neighbors to each patient are evaluated for their relevance, defined by whether they share the progression free survival category on administered therapy types with the case patient. This is then summarized into an average precision score for each patient’s neighborhood, and as a percentage of neighborhoods defined as useful for the whole cohort:
Ideally, we will achieve patient similarity artificial intelligence for real-time learning:
Finally, Dr. Van Allen discussed the How? of integrating artificial intelligence. This may include artificial intelligence ready cancer hospitals, multiple cancer alliances sharing artificial intelligence platforms, and US Senate artificial intelligence and cancer briefings:
Dr. Van Allen concluded his presentation by addressing the big question of: Where do we go from here? To answer this question, he posed two opposing quotes:
- ChatGPT and the Meaning of Life, Harvey Lederman, DPhil “For the last two and a half years, since the release of ChatGPT, I’ve been suffering from fits of dread. It’s not every minute, or even every day, but maybe once a week, I’m hit by it – slack jawed, staring into the middle distance – frozen by the prospect that someday, maybe pretty soon, everyone will lose their job.”
- Stop Worrying, and Let A.I. Help Save Your Life, Robert Wachter, MD “But as the saying goes, ‘Don’t compare me to the Almighty; compare me to the alternative.’ In health care, the alternative is a system that fails too many patients, costs too much, and frustrates everyone it touches. A.I. won’t fix all of that, but it’s already fixed some of it – and that’s worth celebrating.”
Presented by: Eliezer M. Van Allen, MD, Dana Farber Cancer Institute, Boston, MA
Written by: Zachary Klaassen, MD, MSc – Urologic Oncologist, Associate Professor of Urology, Georgia Cancer Center, Wellstar MCG Health, @zklaassen_md on Twitter during the 2026 American Society of Clinical Oncology Genitourinary (ASCO GU) cancers symposium held in San Francisco, CA, between February 26th and 28th, 2026.
Related content: Reimagining Cancer Care: The Role of Artificial Intelligence in Clinical Delivery - Eliezer Van Allen
References:
- Elmarakeby HA, Hwang J, Arafeh R, et al. Biologically informed deep neural network for prostate cancer discovery. Nature. 2021 Oct;598(7880):348-352.
- Shady M, Reardon B, Jiang S, et al. Learning patient similarity from genomics for precision oncology. medRxiv. 2025 Dec 18:2025.12.17.25342480.