Interview
a16z Podcast | Move Fast But Don't Break Things (When It Comes to Computational Biology)
- Structural Shift in Pharma: The pharmaceutical industry is undergoing a fundamental transformation from a fully integrated, in-house model to an "unbundled," asset-light structure, mirroring the movie industry's shift from the 1930s studio system to modern syndication and outsourcing models.
- Drivers of Change: This shift is being forced by increased financial pressure and earnings scrutiny, compelling big pharma to outsource traditional cost centers to specialized external providers who can deliver services more efficiently and profitably.
- Computational Infrastructure (Tuzar Model): Startups like Tuzar enable biotech companies to acquire high-value infrastructure (lab resources and compute) on-demand, utilizing a "cloud-like" consumption model where resources are rented for specific durations (e.g., minutes of computation) and released, eliminating the need for massive upfront capital expenditure.
- "Cloud Biology" Definition: This concept involves conducting real-life biological experiments (e.g., animal models, in vitro assays) through fully systematized, robotic automation where the process is defined by computer code rather than manual human labor.
- Reproducibility via Code: By treating biology as programming, "cloud biology" ensures high reproducibility; rerunning an experiment is functionally equivalent to rerunning code, addressing the historical crisis of irreproducibility in biological research.
- Specific Case Study: A stealth company has digitized animal model testing (historically performed since the 1950s using human subjectivity), utilizing AWS infrastructure, sensors, and cameras to automate measurements (e.g., rat foot size), resulting in a system that is both more reliable and less expensive.
- Data Silos and Aggregation: A significant barrier to progress is the siloing of data within large pharmaceutical companies, though public repositories (NIH, Array Express, EU) already contain millions of assays; industry trends suggest an eventual move toward data sharing or "run-on-firewall" AI models to unlock insights.
- Payer Leverage: Pharmaceutical companies face growing pressure to share data because payers (insurers) are aggregating real-world data and developing analytical capabilities that may exceed the internal data insights of drug manufacturers.
- Market Disruption and Margins: The era of 80% gross margins and double-digit growth is ending; increased R&D risk and shareholder pressure for returns have made the traditional "big pharma" buyout model less viable, favoring smaller, specialized virtual pharma firms.
- Emerging Business Models: New models are emerging, such as the Cystic Fibrosis Foundation's role in funding and recruiting for clinical trials, and the potential for higher profitability in orphan drug trials due to smaller patient pools.
- Core Competency Identification: There is significant internal confusion regarding pharma's core competency, but the consensus is that the translational stage from late-stage assets to commercialization (understanding market needs) remains the hardest-to-disrupt function.
- Diagnostic-Therapeutic Convergence: Advances in genomics, proteomics, and metabolomics are creating opportunities to interface diagnostics with therapeutics, aiming to identify subpopulations that can be treated effectively with existing drugs.
- Future of Personalized Medicine: The long-term goal is a pharmacy system where biological samples are collected and a tailored solution is formulated on-demand; current limitations in cost and data volume are being overcome by analyzing larger populations to build predictive models for subpopulations.
- Reduction of Animal Testing: Machine learning simulations are predicted to eventually replace animal models (which have poor predictive power for humans, e.g., "curing cancer in mice a million times") for clinical trials, a transition that will require significant regulatory (FDA) evolution.
- Cultural Friction: Historically, a deep cultural divide existed between Silicon Valley (computational, "move fast") and Pharma (quality-control focused, slow to digitize), though this gap is narrowing as disruptors from outside the industry increasingly drive innovation.
- Regulatory Lag: Regulatory bodies like the FDA have been slow to adapt, with clinical data only recently allowed to be entered directly into computers rather than paper, reflecting a cautious but necessary focus on data integrity.
- Investment Thesis: Jeff Kindler notes that while large incumbents that successfully adopt virtualization and outsourcing will survive, the most significant innovation will likely originate from new entrants and virtual pharma companies that lack legacy infrastructure.