Building on a Legacy of Actionable Data Science in Medicine
The Late Atul Butte's team share updates, reflecting a unique culture that will continue to set them apart in solving some of the world's most vexing medical problems through data science
AI in medicine is a powerful frontier because it directly shapes patient outcomes, and UCSF’s computational health data scientists are leading the way in developing and applying AI responsibly. On October 31, 2025, the team gathered—wearing Star Trek–themed Halloween costumes—to honor the late Dr. Atul Butte, whose mentorship shaped their work, and to share new research during the monthly UCSF AI Seminar Series. The event was hosted by BCHSI, DoC-IT, the UC Berkeley–UCSF Computational Precision Health Program, and UCSF Epidemiology & Biostatistics.
Key Takeaways
Start with the question — meaningful AI begins with the problem, not the tool.
Validate with a clinical simulation — models must prove their value before implementation.
Navigate thoughtfully — success requires diligence, awareness, and collaboration to avoid misinformed or premature applications of AI.
A Celebration of Vision, Collaboration, and Impact
The October UCSF AI Seminar brought together faculty, clinicians, and data scientists to honor the legacy of Dr. Atul Butte, whose pioneering work defined UCSF’s modern approach to data-driven medicine.
Held on Halloween, the event blended reflection and good humor: attendees came dressed in Star Wars and other costumes, while a photo of Atul—remembered fondly as a Star Trek fan—set a warm, personal tone.
The seminar showcased the extraordinary progress being made in applying AI to two of medicine’s greatest challenges: Alzheimer’s disease and cancer. Featured speakers Marina Sirota, PhD, and Julian Hong, MD, MS—both protégés of Dr. Butte—shared how UCSF continues to build on his vision, bridging computation, collaboration, and clinical impact.
“You could fill a couple of hours talking about the problems of AI in medicine,” one participant noted, “but the key is to do the best work possible and think systematically about every step.”
The seminar audience quickly learned that this is the team to follow when it comes to launching AI innovations in medicine, balancing both first-of-their-kind technologies with what is most efficient, effective, safe and trustworthy for clinicians and patients alike.
Using AI to Tackle Alzheimer’s Disease
Marina Sirota, PhD, opened the seminar with “From Data to Knowledge: Integrating Clinical and Molecular Data for Predictive Medicine.”
She discussed how AI and molecular data can be used to better understand and treat Alzheimer’s disease (AD)—a condition that remains difficult to diagnose early and challenging to treat effectively.
Her research combines computational biology, AI, and large-scale molecular data to uncover new treatment targets and accelerate precision medicine approaches. By integrating real-world clinical data and knowledge networks, her team is helping bridge the gap between discovery and care.
Marina reflected that Dr. Butte’s mentorship taught her to always start with a well-formed question—not with the methods or tools. Sometimes, she observed, “the solution is simple and doesn’t necessarily require data science.”
Applying AI to Help Cancer Patients and Advance Care
Julian Hong, MD, MS, followed with “The Future of Evidence-Based Medicine is Data-Driven Medicine: Leveraging Data and AI in the Cancer Clinic.”
He described how his team uses real-world data and AI tools to improve cancer care, work that became especially urgent during and after the COVID-19 pandemic.
Julian’s research builds on Dr. Butte’s vision of data-driven medicine—applying AI-guided strategies to reduce treatment-related toxicities and improve patient outcomes. His team leverages UCSF’s Information Commons, Wynton, and clinical NLP tools to create actionable predictive models that directly inform care.
He also shared updates from Weill Cancer Hub West, which connects research insights with clinical decision-making—demonstrating how AI can move from theory to impact.
AI in Medicine: Promise, Challenges, and Actionability
The discussion underscored both the promise and the complexity of using AI responsibly in healthcare. Speakers acknowledged the challenge of actionability—ensuring that AI-generated insights are not only predictive but useful for patients and clinicians.
For example, if a model predicts a high risk of Alzheimer’s disease, what can be done with that information? The next steps might include enrolling in a clinical trial or pursuing non-therapeutic interventions, underscoring that AI’s value depends on connecting data to decisions.
Both Marina and Julian emphasized the need for researchers to be aware of potential pitfalls in implementing AI in medicine. Their shared message: the path to success requires a systematic, thoughtful, and collaborative approach. These speakers and other members of the team, many of whom were in the audience, shared a profound understanding of their craft, demonstrating the criticality of their expertise in both technical innovation a deep understanding and consideration of the long-term impact on people and society.
Validation: The Core of Responsible Innovation
Sharat Israni, Executive Director of BCHSI, highlighted validation as one of Dr. Butte’s defining principles.
Atul was an early adopter of new technologies, but he always insisted on rigorous validation, especially in clinical settings. This insistence on testing and proving models before implementation continues to distinguish AI in medicine from other applications of AI—ensuring that innovation remains both ethical and impactful.
“Atul was in the early adoption of new technologies coming out of research, but he insisted you validate early as well. This was a holy thing. If you see it coming out, validate it up front.”
Continuing the Mission
Throughout the seminar, speakers reflected on Atul’s enduring influence—his curiosity, rigor, and focus on questions that matter.
They agreed that while many in this field are researchers rather than clinicians, their shared goal remains to make their work useful for the patient.
We are at a pivotal moment in the history of medicine, as revolutions in biology and immunology converge with advances in AI and data science. Dr. Butte called this a “revolution in health care.”
Thanks to the talented team Atul put in place, UCSF is poised to make measurable differences in the lives of patients and families facing Alzheimer’s disease, cancer, and many other complex conditions, thanks to combined advances in AI, data science and molecular biology
Background reading
Please also reference the event recording: October Special Edition AI Seminar: Marina Sirota and Julian Hong - Honoring the Legacy of Atul Butte, Bakar Computational Health Sciences Institute, October 31, 2025 [RECORDING]
October 31 Seminar Series: Implementation and Evaluation of AI in Real-World Clinical Settings - The October AI seminar will honor Atul Butte's enduring legacy by showcasing projects built on the computational ecosystem he created at UCSF, including high-impact and widely used tools such as the Information Commons, Wynton, PatientExploreR, SPOKE, and clinical NLP services such as cTAKES-as-a-Service and de-identification pipelines.
The UCSF AI Seminar Series is a monthly forum for anyone interested in the intersection of AI and healthcare. See past events and learn more here → UCSF AI Seminar Series
Learn more about UCSF BCHSI → UCSF Bakar Computational Health Sciences Institute website
Gift Launches $200M Initiative for the Weill Cancer Hub West , UCSF News, July 24, 2025. BCHSI affiliated faculty Julian Hong’s Project IMPACT is awarded to build artificial intelligence systems that integrate clinical and biological data for precise, personalized cancer treatments.
[Cover Photo: Attendees in Halloween costumes, with a tribute image of Dr. Atul Butte, a devoted Star Trek fan. From the left, Drs. Peter Washington, Angela Rizk-Jackson, Marina Sirota, Andrew Bishara, Julian Hong, Jean Feng]
Other photos featuring presenters and members of the BCHSI team can be found below.
