Genprex and Roche are collaborating to test whether the protein TROP2 can predict which lung cancer patients respond to Reqorsa, a gene therapy designed to restore a missing tumor‑suppressor gene. Reqorsa does not target TROP2. Roche’s artificial intelligence will score TROP2 on digitized tumor slides to provide a consistent, reproducible measurement that a clinical trial can use.
What the AI does—and what it does not
Roche’s AI is not discovering the biological mechanism linking TROP2 to response. Instead, it converts a previously inconsistent, visually observed marker into a numeric score across digitized histology slides. That reproducible readout allows investigators to test statistically whether TROP2 levels correlate with clinical response to Reqorsa.
A shift in the research sequence
Traditionally, researchers seek to explain a biomarker’s biological role before validating its predictive value in trials. In this case, AI enables validation to come first: the signal is measured and tested in a clinical setting, and mechanistic explanations can follow. This approach does not abandon biological investigation but changes the order of operations.
What researchers are not claiming
According to the available information, there is no confirmed mechanism explaining why TROP2 might predict response to Reqorsa—only hypotheses. The AI does not replace the need for biological experiments; it provides a practical tool to make the signal usable in a clinical trial.
Why this matters
If the trial shows that a TROP2 score predicts therapeutic response, it could open a pathway for testing other subtle or rare histological signals without first fully understanding their biology. That could speed up the validation and clinical use of new predictive markers in personalized medicine.
