Builders Unscripted: Ep. 5 - Derya Unutmaz

Derya Unutmaz, an immunologist, shares how AI, particularly Codex, is transforming biological research, from building custom cell analysis tools and T-cell simulators to envisioning a future of digital twins and personalized medicine, highlighting a coming "Science 2.0" paradigm.

July 6, 2026 7 min read

Builders Unscripted: Ep. 5 - Derya Unutmaz — source: https://www.youtube.com/watch?v=4sexN3yE8xg

AI’s Transformative Power in Biology: A Vision for Science 2.0

Derya Unutmaz, M.D., an immunologist and professor at The Jackson Laboratory, shares his journey and vision for integrating Artificial Intelligence into biological research, from personal productivity to the development of “digital twins,” highlighting a coming “Science 2.0” paradigm.

The Inevitability of AI in Biology

Dr. Unutmaz realized the necessity of AI in biology shortly after graduating medical school in the early 90s, confronted by the “incredible complexity” of biological systems—trillions of parts and billions of reactions. His early interest in AI coding evolved with the deep learning revolution, particularly after AlphaFold and ChatGPT showcased AI’s capability to process massive information in parallel.

The turning point came with OpenAI’s o1-preview reasoning model. Dr. Unutmaz, active on X (formerly Twitter) about AI’s potential, was invited to test the model. He posed a complex immunology question, using a “battle royale” game analogy for the immune system fighting tumors: “How would you develop a scenario where the immune cells would fight against cancer?” The o1-preview’s “deep, insightful response” was emotional and transformative, signaling that AI had reached a point where it could reason, making it “extremely relevant for science.” This moment solidified his belief in AI’s inevitability for scientific advancement, moving beyond literature search and mundane tasks to actual scientific reasoning and prediction.

Codex: The Scientist’s Daily Driver

Dr. Unutmaz describes himself as a “Codex addict,” starting his days by bringing new ideas to life with the AI tool. What previously took “weeks or months” for even experienced coders can now be rapidly prototyped or built. He emphasizes how Codex has become indispensable for generating apps and simulations directly relevant to his research.

Practical Applications of Codex in Immunology:

Flow Cytometry Data Analyzer:
- Challenge: Analyzing complex flow cytometry data, which serves as a “window to the world of cells” (e.g., immune cells), identifying hundreds of cell types based on fluorescent markers. Traditional software is sophisticated but often rigid.
- Codex Solution: Dr. Unutmaz built a custom, fully functional application entirely with Codex, capable of visualizing individual cells (dots on a plot), selecting fluorescent markers (e.g., CD4, CD8 for T cells), applying “gates” to define cell populations, and generating statistical analyses. The app can rapidly process “hundreds of thousands of data points,” demonstrating remarkable optimization and speed.
- Panel Design: An extension allows users to select desired cell types (e.g., naive T cells, TH17 cells) and receive recommendations for optimal marker panels, streamlining experimental design.
T-Cell Receptor Simulation:
- Vision: To move from physical experimentation to digital simulation, much like aerospace engineering, with the ultimate goal of “virtual cells,” tissues, and eventually a “digital twin” of a complete human being.
- Codex Solution: He developed a simulator for the T-cell receptor, a single but “extremely complex” receptor on T cells that dictates crucial immune outcomes (e.g., fighting cancer, autoimmunity, viral infections). The simulator allows manipulation of variables like ligand affinity, dose, and inhibitory molecules to predict molecular activation and phosphorylation patterns. This application allows for rapid in silico testing of hypotheses, such as the effect of a small molecule inhibitor.
Interactive Immune Atlas (ImageGen + Codex):
- Inspiration: Impressed by Image 2.0, Dr. Unutmaz asked it to create a Nature cover based on a published immune atlas.
- Codex Integration: He then fed this generated image to Codex with the prompt: “I want you to understand this image and create an interactive simulator website.” Codex autonomously built a website with “beautiful animations” showing cell interactions. Users can click on cells (e.g., memory cells, effector cells) to get information and even simulate effects like blocking molecules (e.g., PD-1, crucial in cancer therapy), or how immune cells behave based on age. This highlights the potential for advanced educational and exploratory tools.

CRISPR Design Tool:
- Concept: Viewing cells as “programmed code” and envisioning a “Codex for biology” to program cells. CRISPR technology allows genetic engineering, but target selection is complex.
- Codex Solution: Dr. Unutmaz created a native macOS app (with plans for an iPad version) that pulls gene sequences from databases (e.g., CD4 gene), identifies potential CRISPR target regions, ranks them by specificity and predicted effect, and allows users to select and visualize targets. A “design library” feature enables the generation of multiple CRISPR molecules for various genes, dramatically accelerating experimental design.

The Digital Twin and the Future of Medicine

Dr. Unutmaz sees the “digital twin”—a complete biological system of a person—as the pinnacle of AI’s impact. This involves modeling genetics, metabolites, gut microbiome, hormones, and the immune system.

Vision for Personalized Medicine:

Prediction and Prevention: Predicting disease onset and individual responses to treatments.
Tailored Treatments: Moving beyond “one-size-fits-all” drugs (like statins) to highly personalized therapies, where AI simulates a patient’s biology to recommend specific treatments.
Accelerated Clinical Trials: AI could perform clinical trials “in 5, 10 days” instead of 5-10 years, drastically speeding up drug development.
Curing Diseases and Extending Lifespan: He believes that within 10-15 years, with superintelligence and increased compute, we could treat all diseases, reverse aging, and significantly extend human lifespan, challenging the “science fiction” label by emphasizing AI’s exponential advancement.
Cancer Treatment Example: For a cancer patient, a digital twin would allow doctors to “try different hypotheses and experiments” in silico, optimizing treatment strategies based on individual mutations, immune responses, and even creating personalized RNA vaccines, as seen in recent groundbreaking cases.

The Science 2.0 Paradigm Shift

The way science is conducted is undergoing a “complete fundamental shift.” The traditional model of weeks of planning and months of analysis is “over.”

AI-driven Hypothesis Generation: AI agents will generate hypotheses, simulate experiments, and analyze data almost instantly.
Automated Labs: Human scientists will direct AI agents to “go figure this out,” with automated labs and robots performing experiments.
Jevons Paradox: The ability to do more will lead to doing “so much more,” fostering new fields like bioengineering to create new cell types and tissues.
Broad Impact: This transformation extends beyond biology to physics, material sciences, and chemistry (e.g., drug discovery, clinical trials).

Embracing the AI Age: Agency and Curiosity

Dr. Unutmaz’s advice stems from his experience with constant experimentation and failure in biology: “agency and curiosity are the only things that matter.”

Experiment Constantly: The “cost of experimentation is so low” with AI tools.
Overcome Fear: Don’t be scared of the new; embrace it.
Positive Outlook: AI is not a threat but a force that will “enhance us to a golden age,” making AI researchers “heroes” for driving “the greatest transformation of humanity.”