AI Transforming Health: AlphaFold and Beyond | The Check Up with Google ‘25

- Intersection of Health and Technology [00:10]: The speaker highlights their career at the intersection of health and technology, including medical imaging, digital pathology, and surgical robotics, and emphasizes AI as a transformative technology with immense potential for health.
- AlphaFold: Predicting Protein Structures [00:36]:
  - What it is: AlphaFold is an AI technology designed to predict the 3D structure of proteins.
  - Why it’s important: Proteins are the fundamental building blocks of life, and their function largely depends on their structure. Mapping these structures was previously a very complex and time-consuming process (e.g., an entire PhD for a single protein) [00:56].
  - Impact: AlphaFold has predicted over 200 million proteins, almost all proteins known to science. These predictions have been released to the scientific community in a freely accessible database in collaboration with the European Molecular Biology Laboratory (EMBL-EBI) [01:27].
  - Research Acceleration: Over 2.5 million scientists in more than 190 countries have used AlphaFold to accelerate research in areas like fighting malaria, advancing research on neglected diseases, and deepening the understanding of tuberculosis for potential new treatments [01:50].
- The Future of Health: Drug Discovery with TXGemma [02:28]:
  - Challenge in Drug Discovery: Traditionally, discovering and developing new medicines is complicated, costly, and burdensome.
  - TXGemma Introduction: Google is announcing TXGemma, a collection of open models designed to improve the efficiency of AI-powered drug discovery [02:49].
  - Built from Gemma: TXGemma is built from Gemma, a family of lightweight open models derived from the same research and technology used to create Gemini models [03:04].
  - How TXGemma Works: It understands both regular text and the structures of therapeutic entities like small molecules, chemicals, and proteins. Researchers can ask TXGemma questions to predict properties of potential new therapies, such as their safety or effectiveness [03:12].
  - Availability: TXGemma will be released as part of the Health AI Developer Foundations (a collection of models and resources) later this month, making it directly accessible to developers and researchers [03:39].
  - Performance Evaluation: TXGemma was evaluated on over 60 key tasks in the drug discovery pipeline, including gene-disease association, drug target identification, and prediction of clinical trial outcomes [04:03].
  - Real-World Potential (Ovarian Cancer Example) [04:22]: The model successfully identified a gene associated with ovarian cancer and a specific drug that could block it, even though neither the gene nor the drug was explicitly linked to ovarian cancer in the training data. This drug has since shown activity against ovarian cancer in current medical research, highlighting TXGemma’s predictive power.
  - Interactive Conversational Mode: Researchers will be able to interact with TXGemma in a conversational mode, asking follow-up questions and understanding the reasoning behind the model’s predictions, moving beyond single-answer responses [05:24].
- Conclusion [05:47]: AlphaFold and TXGemma represent a new generation of AI technology impacting health at different scales, with the potential to translate research into improved patient care. Google aims to continue collaborating with partners and the scientific community.