MD Anderson Uses AI to Track Cancer Treatment Resistance

Researchers employed an RNA foundation model to map biological states linked to drug resistance in metastatic colorectal cancer.

Updated on Oct. 22, 2026 in Cancer

MD Anderson Uses AI to Track Cancer Treatment Resistance

Live Poll

Should medical researchers prioritize using AI-driven biological modeling over traditional clinical response signatures?

A team at the University of Texas MD Anderson Cancer Center has utilized the Eigen-1 RNA foundation model to identify transcriptional states in patients with metastatic colorectal cancer. This unsupervised AI approach successfully detected biological mechanisms of treatment resistance without relying on predefined clinical response signatures.

Why it matters

Understanding how colorectal cancer resists specific therapies is essential for developing more effective treatment regimens for patients. These findings offer a new method for uncovering the biological processes that cause patients to stop responding to standard cancer interventions.

The study recorded an 81% post-hoc concordance between transcriptional states and RECIST-defined clinical responses. This follows a prior phase 1/2 study that observed a 50% objective response rate for a drug combination of encorafenib, cetuximab, and nivolumab.

The players

University of Texas MD Anderson

This is a world-renowned cancer treatment and research institution based in Houston that leads major clinical oncology studies.

Eigen Bio

This organization specializes in the development of computational biology tools and RNA foundation models for medical research.

Van Morris

He is an oncology researcher who previously led a phase 1/2 study investigating combined therapy for metastatic colorectal cancer patients.

The details

Researchers applied the Eigen-1 model to pretreatment and on-treatment blood samples from patients with MSS BRAF V600E metastatic colorectal cancer. By mapping these findings to mechanistic networks, the team generated hypotheses regarding how cancer cells escape treatment, distinct from results found in the SWOG S2107 trial where adding nivolumab failed to improve survival.

Study results were published on October 22, 2026.
Findings are scheduled for presentation at the European Society for Medical Oncology Congress in 2026.

The Big Picture

This research provides a new analytical perspective on patient resistance that directly contrasts with the findings of the SWOG S2107 trial, which failed to show survival improvements with specific drug additions. It marks a shift toward using unsupervised machine learning to discover treatment escape mechanisms rather than relying on historical clinical labels.

For patients undergoing cancer treatment, this research represents a potential future improvement in how doctors determine which therapies will be effective for specific biological profiles. It does not currently change existing treatment protocols but points toward more personalized medicine strategies in the future.

The takeaway

AI foundation models are increasingly becoming powerful tools for identifying biological patterns that were previously invisible in cancer research. Patients and providers should watch for follow-up studies to see if these models can successfully predict therapy outcomes in larger groups.

What happens next

The research team plans to conduct future work to determine if the biological states identified by the Eigen-1 model recur in additional patient groups and different clinical treatment settings.