Decoding the Tumor: Inside the AI-Driven Hunt for Colorectal Cancer Immunotherapy

Colorectal cancer (including colon cancer and rectal cancer) is the second leading cause of cancer-related death in the United States, and is now the leading cause of cancer death among adults under 50. Despite decades of progress, immunotherapy – one of the most powerful tools in modern oncology – still falls short for the vast majority of patients. Karin Pelka, PhD, an investigator at The J. David Gladstone Institutes and a Cancer Research Institute (CRI) Technology Impact Award grantee, is working to change that by listening in on the molecular conversations happening inside the tumor itself.

Addressing the Immunotherapy Gap in Colorectal Cancer

Immunotherapy works best when the immune system can already recognize a tumor as foreign. Some cancers – like melanoma or lung cancer – can accumulate enough DNA mutations that they look distinctly “foreign”, triggering a strong immune response that checkpoint inhibitors can amplify. However, most colorectal cancers don’t work that way.

The vast majority of colon cancer falls into this relatively large area of tumors where we haven’t quite figured out yet how to use the immune system, really, to fight them.

There is one critical exception: roughly 5–10% percent of patients with colorectal cancer carry tumors with a DNA mismatch repair defect (dMMR), causing them to look immunologically distinct. For these patients, the results have been extraordinary: a 2022 trial showed that locally advanced rectal cancer patients in this group, who previously would have undergone chemotherapy, radiation, and surgery, could be treated with immunotherapy alone.  Every patient responded in that trial, and there was no detectable cancer remaining. 

Dr. Pelka’s mission is to extend that kind of extraordinary outcome to the other 90% of patients.

Screening Matters!
Colorectal cancer can often be prevented, not just detected. A colonoscopy can find and remove polyps before they develop into cancer–a process that can take 10 to 15 years–making regular screening starting at age 45 one of the most powerful tools available.

Mapping Hotspots Inside the Tumor

A key insight from the Pelka Lab points toward how it might be possible to target challenging colorectal cancer. Her team identified structured clusters of immune activity within tumors called “immune hubs” where killer T cells, cancer cells, and the surrounding structural tissue engage in intense signaling. Whether that signaling helps destroy the cancer or allows it to grow depends on the dynamics of the communication cascade.

Interestingly, the presence of immune hubs in a particular activated state turned out to be highly predictive of immunotherapy response – not just in colorectal cancer, but across varying tumor types, including melanoma and lung cancer. The data suggest that these hubs are conserved and exploitable across cancer types, and the question is what’s preventing the hubs from forming, or staying active, in patients who currently don’t respond to immunotherapy.

Teaching AI to Read the Cell

Finding the answer requires making sense of an enormous amount of biological data, which is where Dr. Pelka’s CRI Tech Impact award is enabling genuinely new science. Her lab is applying Geneformer, a foundational AI model developed by collaborator Dr. Christina Theodoris, originally trained on tens of millions of human cell profiles to recognize patterns in gene activity, akin to the way a language model learns patterns in text.

The goal is to identify the master molecular switches, known as “central regulators,” that control whether cancer cells are visible to the immune system, and to use spatial transcriptomics to map how those cells interact across the tumor’s physical landscape. Together, these tools could surface new immunotherapy targets that work across many patients, not just a genetically defined subpopulation.

Why This Work Can’t Wait

For patients diagnosed in young adulthood – the population now driving the rise in colorectal cancer diagnoses – five more years is not enough. Targeted therapies, while precise, face a fundamental limitation: cancer cells evolve around them. Immunotherapy, when it works, aims to generate what is often called a “living therapeutic,” one that can adapt alongside the cancer and keep it in check long-term. That is the possibility that draws researchers like Dr. Pelka to this work — not just more time, but the future patients were planning on.