Story of a LLM Adjacent Daraxonrasib Trial
OpenAI deep research was utilized in a June 2025 work featuring 40 Pancreatic Ductal Adenocarcinoma (PDAC) meta-analyses totaling over 400,000 words, many based on clinical trials. Gemini ranked a daraxonrasib drug combination as the top candidate for ChemicalQDevice to receive funding. The Gemini finding then served as an input for OpenAI and Anthropic to generate daraxonrasib-based financial proposals. (1)
An August 2025 AI generated Quantitative Systems Pharmacology (QSP) metastatic pancreatic cancer Python clinical trial simulation was verified as numerically and biologically stable. The median overall survival (mOS) was 12.8 months with the daraxonrasib combination, and 5.4 months with chemotherapy. (2)
The May 2026 Revolution Medicines RASolute 302 trial had similar outcomes. “The median overall survival in the RAS G12 population was 13.2 months with daraxonrasib and 6.6 months with chemotherapy”. (3)
The numerical proximity between the a) QSP Simulation and b) RASolute 302 is an important chronology and hypothesis-supporting observation, not a validation claim because of differences in Sample Size: a) 1000, b) 241; Experimental Drug: a) daraxonrasib combination, b) daraxonrasib; and Molecular Selection: a) KRAS G12C, b) Primarily G12D/G12V.
In 2026, the Physical AI component was added to daraxonrasib-based oncology trial studies using Claude Code. This process required adapting current FDA, ICH, and NIH regulations and templates to a new state of the art AI standard. Resulting documents were then used to draft new legislation, Phase 1 and Phase 2 trial protocols, an Investigational New Drug application, and an NIH “Daraxonrasib Phase 1 LLM-Directed Robotic Whipple in KRAS-Mutated PDAC” trial funding application. (4)
Therefore, from June 2025 to July 2026, LLMs were used to a) Identify the performant pancreatic cancer drug daraxonrasib based on trial literature, b) Generate code that verified the drug’s efficacy versus control, and c) Iterated through the creation of regulatory documents to aid the clinical trial funding application. The 2025 QSP trial had a daraxonrasib combination median overall survival 2.4x increase vs. chemotherapy, similar to the 2026 RASolute 302 daraxonrasib 2x mOS increase vs. chemotherapy.
References:
(1) https://doi.org/10.5281/zenodo.15735068
(2) https://doi.org/10.5281/zenodo.17001137
(3) https://www.naftemporiki.gr/wp-content/uploads/2026/06/NEJMoa2605555.pdf
(4) https://doi.org/10.5281/zenodo.21317266