Why building new proteins from scratch is our new superpower | David Baker

Episode Summary

Executive Summary: David Baker explains how protein design has shifted from slow, physical modeling to powerful AI-driven generation, enabling bespoke proteins for medicine, sustainability, agriculture, and sensing. He argues that designed proteins can outperform evolution in problems nature never faced, that openness and global access matter, and that responsible safeguards should center on DNA synthesis oversight.

Main Topics: From natural proteins to designed proteins (Priority: 5/5): Baker describes why proteins were long considered difficult or nearly impossible to redesign, since natural proteins seemed like highly evolved, almost magical entities with precise functions. The AI transition in protein design (Priority: 5/5): He explains the move from physics-based modeling to AI systems trained on roughly 250,000 known protein structures, enabling generation of entirely new proteins from scratch. Applications in medicine and vaccines (Priority: 5/5): The conversation highlights advances in therapeutics, including a COVID vaccine developed from de novo design and the expectation that many more designed medicines will be approved. Sustainability and environmental solutions (Priority: 4/5): Baker emphasizes new protein tools for breaking down plastics, pollutants, PFAS, and for helping remove carbon, reduce methane, and support green chemistry. Agriculture and climate resilience (Priority: 4/5): The team is applying protein stability and design to crops, especially improving thermotolerance in plants like rice and protecting against pests and blights. Technology, sensing, and protein-electronics interfaces (Priority: 3/5): He discusses designing synthetic receptors for artificial noses and embedding proteins in silicon nitride chips to connect biological sensing with electronics. Access, openness, and safety governance (Priority: 5/5): Baker advocates open methods, global-health carve-outs, and DNA-synthesis tracking as the key safeguard against misuse, while noting broader concerns about AI outside biology.

Key Arguments: Protein design was historically hard because scientists lacked both methods and conceptual tools for creating new proteins with new functions. The field progressed from predicting protein structure to working backward from desired function/structure to new sequences and synthetic genes. AI trained on large structure databases now enables generation of novel proteins, similar to prompt-based image generation. Designed proteins can address problems nature never evolved to solve, such as plastic degradation, PFAS breakdown, higher-temperature crops, and some sensing tasks. A de novo-designed COVID vaccine demonstrates that the approach can yield real medicines, not just lab prototypes. Open sharing of tools accelerates scientific progress because other labs can build on them and an ecosystem forms around the methods. Responsible oversight should focus on the DNA synthesis step, with logging and traceability to detect suspicious synthetic sequences. The greatest near-term misuse risks in biology are still those already present in nature; designed proteins are more useful for defense than offense. CRISPR and protein design are complementary: design creates the protein, while gene-editing tools can help deliver it into plants or other systems. Over time, intentional protein design is likely to replace random library screening for antibodies and other biologics because it is more targeted and development-ready.

Data Points: Number of known protein structures used for AI training: about 250,000 - Baker says current AI design models are trained on protein structures determined over the last 50 years. Year of Baker's TED Talk: 2019 - Referenced as the starting point for comparing progress in protein design. Year of the membership conversation: June 11, 2025 - TED notes the date of the conversation with Whitney Pennington Rogers. Former colleagues present at Nobel-related celebration: 185 people - Baker says 185 former colleagues, students, and postdocs came to celebrate in Stockholm. Former trainees at his group now running their own labs: over 100 - He notes more than 100 former graduate students and postdocs have started their own labs.

Pivotal Quotes: "designing new proteins can save the world" — David Baker: Summarizing the motivation behind his work and the premise of the conversation. "We can say, design a protein which binds to this virus and blocks it, or binds to this cancer cell and stops it from dividing." — David Baker: Explaining how RF diffusion works as a prompt-like AI protein generator. "the way to control things and to make sure and Is was to was through the synthetic gene manufacturing step" — David Baker: Discussing biosecurity safeguards and where oversight should be focused.

Implications: Protein design is moving from frontier science to a practical platform for medicines, climate tech, agriculture, and sensing. The big challenge now is scaling access, maintaining openness, and enforcing synthesis-based safeguards so the benefits reach broadly without enabling misuse.