Ep59: David Baker

Episode Summary

Executive Summary: David Baker traces how protein design evolved from understanding natural protein folding to building new proteins from scratch, enabled by Rosetta, distributed computing, cheaper DNA synthesis, and deep biological insight. The conversation highlights his shift from sequence-to-structure prediction to de novo design, the creation of startups and vaccines/therapeutics, and his flat, highly collaborative lab culture.

Main Topics: From curiosity-driven biology to protein folding (Priority: 5/5): Baker recounts how he moved from social studies/philosophy at Harvard into biology after senior-year courses and a fascination with cells, development, and brain biology. Protein folding became his central problem because it was a tractable model of biological self-organization. Rosetta and the sequence-to-structure problem (Priority: 5/5): His early lab work focused on understanding how amino acid sequence determines 3D protein structure. That work helped inspire Rosetta, a computational framework that predicts folding by sampling conformations and scoring energies. Distributed computing and citizen science (Priority: 4/5): To overcome compute limits, Baker’s group built Rosetta@home using dormant home computers and later Foldit, where human players guide folding. These tools expanded the lab’s capacity and public engagement. Transition from prediction to de novo protein design (Priority: 5/5): The lab moved from modifying natural proteins to building entirely new ones from first principles. This shift was driven by the need for better control, broader design space, and the limits of relying on natural scaffolds. Biomedical and industrial applications (Priority: 5/5): The lab and its spinouts are applying design to therapeutics, vaccines, enzymes, cell therapies, mineralization, and materials. Examples include NeoLeukin for IL-2-based cancer immunotherapy and Icosavax for nanoparticle vaccine platforms. Management style and research culture (Priority: 4/5): Baker emphasizes a flat organization, constant conversation, and avoiding hierarchy. He says innovation comes from people working close together, and he stays focused on his own project while enabling broad collaboration. Strategic timing and ecosystem building (Priority: 4/5): He argues the field’s success reflects a confluence of better algorithms, more powerful computing, cheaper gene synthesis, and improved biological understanding. He also wants to build a Seattle-centered biotech ecosystem and the ‘Bell labs of protein design.’

Key Arguments: Protein folding was the right foundational problem because it is a simple, fundamental example of biological self-organization and connects sequence to function. Computing made protein structure prediction feasible by allowing massive sampling of conformational space and energy evaluation. Citizen science was not a side project but a core computational resource: Rosetta@home supplies much of the lab’s compute power. The field had to move beyond natural proteins because evolution samples only a tiny fraction of possible sequences and natural scaffolds impose historical constraints. Designing proteins from scratch enables more precise therapeutics and materials than modifying existing proteins. The best short-term payoff areas are those with well-understood biology, such as IL-2 cancer immunotherapy and vaccine targets like RSV/HIV, because design requires clear specifications. Scientific innovation is accelerated by flat, highly interactive teams where ideas move freely rather than through hierarchy. The current protein design wave is possible because of a confluence of advances: Rosetta, improved hardware, cheap DNA synthesis, and deeper biological knowledge.

Data Points: Years since starting faculty role at University of Washington: 25+ years - Baker says he began his faculty position there at the end of 1993 and has been building the program since 1994. Rosetta@home contributors: many hundreds of thousands of people signed up - He describes the distributed computing project as a major source of computational capacity. Active Foldit players: several hundreds - He says the game has a smaller but engaged steady-state player base. Daily Rosetta@home contributors: upwards of 20,000 to 30,000 - Baker gives a current estimate of active daily contributors. Possible sequences for a 100-residue protein: over 10^130 - He uses this to illustrate the astronomical size of protein design space. Naturally sampled protein sequences on Earth: on the order of 10^15 to 10^20 - Baker estimates the evolutionary search space sampled by life is tiny relative to all possible sequences. Known protein sequences: roughly 10^9 to 10^11 - He notes the small number of proteins actually known compared with the total possible space. First de novo protein design demonstration: 2003 - He cites the first publication showing designed proteins could fold into new structures. Toxicity-reduction goal for IL-2 design: avoid alpha subunit binding; retain beta/gamma binding - The NeoLeukin program aims to preserve anti-tumor signaling while minimizing toxicity. Impact of IL-2 in melanoma: about 10% - The host mentions IL-2 can provoke strong anti-tumor responses in a subset of melanoma patients. Potential buildout grant: $45 million over 5 years - Baker references an audacious project grant supporting expanded in-house capabilities. Company formation rate: about 2 companies/year - He says the institute has been spinning out startups at roughly this pace. Recent company count: about 8 companies in 4 years - Baker describes the recent startup output from the institute.

Pivotal Quotes: "we're creating an entirely new field of chemistry" — Luke Timmerman quoting Raymond Deshaies: Introduces the scientific significance of Baker’s work. "if you really went back to the drawing board and learned how to make proteins systematically from scratch, then our capabilities would expand enormously" — David Baker: Explains the motivation for the shift to de novo protein design. "I don't believe in the scientists on the mountaintop" — David Baker: Describes his flat, collaborative management philosophy.

Implications: Protein design is moving from prediction to engineering, with major implications for immunotherapy, vaccines, enzymes, and materials. The field’s next gains will likely come from startups and partnerships translating academic breakthroughs into real products.