a16z Podcast: The Genetics Of Drug Delivery

Episode Summary

Executive Summary: Stanford professor Russ Altman explains pharmacogenomics and argues that drug response should be studied across molecular, cellular, clinical, and population levels. Using PharmGKB and real-world data, he shows how genetics, adverse-event databases, EMRs, web searches, and social signals can reveal hidden side effects, dangerous interactions, and new uses for old drugs.

Main Topics: Pharmacogenomics and PharmGKB (Priority: 5/5): Altman introduces PharmGKB, a 16-year knowledge base mapping how human genetic variation affects drug response and why inherited variation can determine efficacy and side effects. Codeine as a genetic-response example (Priority: 5/5): He uses codeine metabolism to illustrate how genetic differences in liver enzymes can make a drug ineffective for some people or overly potent for others. Multiscale view of drug response (Priority: 5/5): Altman argues drugs must be understood at molecular, cellular, tissue/organism, and population levels, with data integration increasing confidence in real signals. Mining real-world data for hidden side effects (Priority: 4/5): By combining FDA adverse-event records and electronic medical records, his team can validate known label effects and uncover many additional side effects. Detecting drug-drug interactions (Priority: 5/5): He highlights work showing that drugs individually safe for glucose can cause major hyperglycemia when combined, demonstrating the need to study combinations, especially in older adults. Repurposing old drugs for new indications (Priority: 4/5): Altman explains how side effects, interaction patterns, and binding/genomic clues can suggest new therapeutic uses, especially in oncology. Social and search data as surveillance tools (Priority: 3/5): He describes using web searches, Twitter, Facebook, and patient portals to detect patient-reported symptoms and preferences, despite challenges in language mapping.

Key Arguments: Human drug response is partly inherited just like height, eye color, and other traits, so genetics can guide prescribing. A single drug cannot be fully understood from its approved indication alone; developers often miss broader biological effects. Integrating molecular data with EMRs and population databases reduces noise and increases confidence in findings. Real-world data can reveal dozens or hundreds of side effects not included in the drug label. Older adults often take many medications, making drug-drug interaction studies essential for safety. Signals found in side effects and interactions can point to repurposing opportunities for existing drugs. Social media and search behavior can provide additional, patient-generated evidence of adverse effects and treatment preferences.

Data Points: PharmGKB duration: 16 years - Length of time the knowledge base has been built and expanded. Codeine non-response prevalence in Europeans: 7% - Portion of people of European descent who lack the enzyme variant needed to convert codeine into morphine. Side effects found beyond labels: tens or hundreds per drug - Additional adverse effects identified using FDA records and electronic medical records. Typical medication burden in older adults: 7 to 10 medications - Common medication load for people above 70 taking any medications. Search log analysis result: remarkable increase - Higher occurrence of hyperglycemia-related terms when both paroxetine and pravastatin were searched together. Twitter data scale: tens of thousands of tweets - Volume of tweets that can be mined to identify patient-reported drug side effects.

Pivotal Quotes: "your response to drugs was inherited from mom and dad and grandma and grandpa, just like your height and your hair color and your eye color" — Russ Altman: Explaining the core concept of pharmacogenomics and why genetics matters for prescribing. "we were able to find tens or hundreds of extra side effects per drug with very high confidence" — Russ Altman: Describing the value of combining FDA adverse-event data with electronic medical records. "The average person who's above 70 and who's on any medications is often on 7 to 10 medications" — Russ Altman: Illustrating why drug-drug interactions are a major clinical concern.

Implications: Drug discovery and prescribing will increasingly rely on integrated data science: genetics, EMRs, population data, and patient-generated signals. This could improve safety, personalize therapy, and uncover new uses for existing drugs.