Ep178: Peyton Greenside on ML-Guided Biologics Discovery

Episode Summary

Executive Summary: Peyton Greenside traces her path from math- and biology-focused upbringing in North Carolina to co-founding Big Hat Biosciences, where she argues that the winning formula is pairing rapid cell-free protein synthesis with machine learning to generate better antibody drugs faster. She describes Big Hat’s evolution from platform-building to an oncology/immunology pipeline, upcoming clinic entry, and a CEO transition centered on external execution and clinical validation.

Main Topics: Greenside's background in math, biology, and computational biology (Priority: 5/5): She grew up in Durham/Chapel Hill in a science-heavy family, studied applied math at Harvard, then moved through Cambridge and the Broad Institute before Stanford, ultimately finding computational biology as the bridge between her two interests. The importance of mentorship and encouragement for women in STEM (Priority: 5/5): Greenside discusses being one of the few women in math/computational fields and credits her father, teachers, and later managers for support that helped her persist and thrive. How data quality and scale shaped her scientific worldview (Priority: 5/5): Her Broad and Stanford work reinforced that machine learning in biology depends on high-quality, trustworthy data, and that data generation, QC, and interpretation are as important as modeling. Founding Big Hat around rapid feedback loops in protein design (Priority: 5/5): Big Hat was built to combine cell-free synthesis, synthetic biology, and machine learning so the company could test protein designs quickly, gather rich data, and improve predictive models iteratively. Platform evolution into differentiated therapeutic modalities (Priority: 5/5): The company moved from basic affinity and stability optimization into more advanced engineering such as logic-gated antibodies, pH conditionality, avidity-based designs, ADCs, and T cell engagers for solid tumors. Partnership strategy and capital formation (Priority: 4/5): Big Hat raised venture capital and formed multiple pharma collaborations to broaden disease areas, validate the platform, and help choose the best molecules for development while maintaining a wholly owned pipeline. CEO transition and the path to the clinic (Priority: 4/5): Greenside explains that becoming CEO mostly changes where she spends her time—more external engagement with investors and partners—and that the next major proof point is clinical success, beginning with a lead gastric cancer ADC.

Key Arguments: Computational biology became possible and compelling because falling sequencing costs and expanding data made it feasible to connect genome/epigenome data to disease and phenotype. Machine learning in biology is only as good as the data feeding it; quality control and biological relevance are prerequisites, not afterthoughts. Cell-free protein synthesis is a major enabler because it compresses protein design cycles from weeks to days and supports thousands of designs per week. Feedback speed is the core advantage in protein therapeutics: proteins can be built, measured, and optimized much faster than disease outcomes can be observed in genomics. Big Hat's platform became more valuable as it learned to engineer not just binding, but higher-order functions like gating and conditional targeting. The company intentionally focuses on engineering risk it can control rather than chasing entirely novel biology, because that path is more likely to reach patients. Partnerships with major pharma validate the technology, broaden the company’s scientific exposure, and help select the best molecules for clinical advancement. Clinical proof will ultimately determine whether AI-plus-synthetic-biology platforms can improve the probability of success, not just the speed of discovery.

Data Points: Big Hat founded: 2019 - Company start date given in the introduction and discussed throughout the interview. Venture capital raised: more than $100 million - Described in the introduction as funding secured by Big Hat. Pharma partnerships mentioned: AbbVie, Johnson & Johnson, Amgen, Merck, Eli Lilly - Greenside lists major collaborators and notes Lilly as the newest partnership. Harvard attendance: 2007 to 2011 - Greenside states her undergraduate years at Harvard studying applied math. Protein production timeline in cell-free vs. traditional systems: days vs. about 4 weeks - She contrasts cell-free synthesis with conventional cell-based expression workflows. Throughput in cell-free platform: thousands of unique antibody designs per week - Greenside describes the scale of Big Hat's current platform output. Scale improvement in data matrix: two orders of magnitude larger - She says the platform can create a far larger sequence-by-property matrix than typical workflows. First proof-of-concept cell-free run: 32 antibodies - Early experiment validating binding in cell-free synthesis. Initial model training set size: a few hundred antibodies - She says the first machine learning models were trained on only a few hundred examples. Early program timing: founded about six months before COVID - Used to explain why the company initially explored COVID antibodies. Lead program timeline: entering the clinic in 2026 - Greenside says the lead gastric cancer ADC is expected to enter first-in-human studies next year. First molecule development time: 9 months - Example of an avidity-based cancer-targeting molecule design cycle. Second version development time: 3 months - She cites the faster timeline for a follow-on design against a new target.

Pivotal Quotes: "You know, I think the sort of themes that I've worked on so far have kind of led me closer and closer to that paradigm. I'm really believing firmly in that feedback loop." — Peyton Greenside: Explaining why she gravitated toward protein design and rapid experimental feedback over slower genomics problems. "At least 90% of machine learning and biology in a lot of ways is cleaning and understanding your data and ensuring that the signal there, again, reflects what you're trying to learn." — Peyton Greenside: Describing the central role of data quality in computational biology and model performance. "We can encode logic in our therapeutics, and that was not clear to us, you know, in the founding." — Peyton Greenside: Discussing how Big Hat's platform evolved from basic optimization into advanced antibody logic for oncology.

Implications: The interview suggests AI biology wins when paired with fast experimental cycles, not just better models. For biotech, the near-term value may be in faster, more reliable optimization of therapeutics, with clinical validation determining whether platform-driven drug design truly improves success rates.