Vijay Pande: Why Biology Is Now an Engineering Science
Vijay Pande, known for Folding@home and building a16z's biotech practice to nearly $4 billion, made a surprising pivot last year. He left to co-found VZVC, a tiny, AI-driven firm making only about five concentrated bets a year. In a recent interview, Pande explained why biology is becoming an engineering discipline, the persistent cost of clinical trials, and why open, shared datasets—not walled-off ones—are key to AI transforming medicine.
From Discovery to Engineering
Pande argues that biology is moving from a 'science of discovery' to something you can engineer. Historically, drug development had a fortuitous aspect, but AI and machine learning now allow computers to understand complex biological systems, identify drug targets, design drugs, and even assist in clinical trials. This shift is not just theoretical; it's practical. AI models are not perfect, but they are already better than animal models at predicting human responses, which is why drugs often fail in trials. Once AI crosses that bar, the potential for more effective and cheaper drug development becomes real.
The Persistent Cost of Clinical Trials
Despite hopes that synthetic data would reduce trial costs, Pande notes that clinical trials remain brutally expensive, often costing hundreds of millions of dollars. The probability of a drug succeeding from first to third trial is just 20%, meaning eight out of ten fail. This high failure rate, often due to poor predictive power of animal models, drives up amortized costs. Pande sees AI as a way to improve the odds, making trials more efficient and potentially reducing costs over time, but the aspiration of cheap trials is still far from reality.
The Data Dilemma: Open vs. Walled Gardens
One of the most interesting conundrums in AI-driven biotech is that biological data can't be scraped off the internet like text. This leads to companies building proprietary datasets, creating silos. Pande acknowledges the tension: founders and investors want to protect their findings, but he sees a shift toward building 'atlases' of biological information, often as foundation models. He predicts that, like open-source LLMs, open-source foundation models in biology will have a broad impact. This would enable AI to act as a 'specialist in everything,' potentially integrating knowledge across specialties and improving patient care.
Key Takeaways
- Biology is becoming an engineering science, with AI enabling more targeted drug design.
- Clinical trials remain expensive and failure-prone, but AI could improve success rates.
- Open, shared datasets are crucial for AI to reach its full potential in medicine.
- VZVC is intentionally small, making only about five concentrated bets per year.
- Go-to-market is as hard as technology, a lesson Pande emphasizes to founders.
Source: TechCrunch • 🇺🇸 San Francisco
Keep Reading
