Conference Presentation, Lecture, Keynote
a16z Podcast | The Genetics Of Drug Delivery
- Russ Altman, a Stanford professor of bioengineering, genetics, and medicine, leads a laboratory focused on biomedical informatics and data science to understand and optimize drug response.
- Altman's primary funding source is the National Institutes of Health, with active collaborations including Pfizer, Genentech, Karius, and Second Om.
- Altman is a founder of Personalis, an immuno-oncology company, and has spent the last 16 years developing the Pharmacogenomics Knowledge Base (PharmGKB).
- PharmGKB serves as a knowledge base documenting how human genetic variation impacts drug response, leveraging the fact that drug metabolism and efficacy are inherited traits similar to physical characteristics.
- A specific case study involves codeine: 7% of people of European descent lack a functional enzyme to convert the biologically inactive codeine into active morphine, rendering the drug a placebo for this demographic.
- Conversely, some individuals metabolize codeine into morphine so rapidly they may experience intense pain relief followed by a prolonged gap until the next dose, highlighting the need for genetic screening to prevent toxicity or under-dosing.
- Altman argues that current pharmaceutical development creates a fragmented view of drugs because trials are designed with narrow "blinders" focused on proving specific intended effects rather than capturing the full spectrum of drug activity.
- To correct this, Altman's lab integrates data across four biological scales: molecular interactions, cellular responses, tissue/organism levels (using electronic medical records and wearables), and population-level databases.
- By cross-referencing signals across these scales, the lab can distinguish genuine biological effects from data noise, a capability Altman believes is essential for the next generation of drug discovery.
- Using FDA adverse event reports combined with electronic medical records, the lab successfully replicated official drug labels while identifying tens to hundreds of additional side effects per drug with high confidence.
- This multi-scale analysis allows researchers to differentiate between class effects (side effects common to a drug family) and drug-specific effects, which is critical for market positioning and safety profiles.
- The lab identified a previously unreported drug interaction between paroxetine (Paxil) and pravastatin; while individually neutral regarding blood glucose, their combination caused a significant increase in serum glucose, particularly in diabetics.
- This discovery was validated by analyzing population data, electronic medical records, and mouse organism-level data, followed by a search log analysis with Microsoft showing a marked increase in patient queries regarding hyperglycemia symptoms when both drugs were mentioned.
- Altman proposes using direct patient surveillance via web search logs, social media (Twitter), and Facebook patient portals to identify side effects and gauge patient preferences, noting the challenge of mapping informal language to medical concepts.
- The lab is applying these insights to drug repurposing, suggesting that new indications for existing drugs can be discovered by analyzing side effect profiles and unintended molecular binding patterns rather than approved label indications.
- Specific applications include predicting that cancer drugs approved for one indication may be effective against different cancer types based on genome data and binding patterns, providing a basis for clinical trials or off-label use.
- Altman expresses optimism that data science, specifically the integration of diverse data streams, will optimize future drug discovery and allow for continuous monitoring of drug performance once exposed to patients.