Daphne Koller: How Machine Learning Is Revolutionising Drug Discovery | WIRED Health

AI's Transformative Role in Drug Discovery

• 🚀 Machine learning is accelerating every step of the drug research and development (R&D) value chain, from discovering therapeutic hypotheses to identifying chemical matter and optimizing clinical development.
• 💡 The goal is to use AI to unlock and derisk novel therapeutic hypotheses, rapidly identify chemical matter with desired properties, and pinpoint the right patients and endpoints for more efficient clinical trials.

Addressing Drug Development Challenges

• ⚠️ The drug industry faces a success rate of less than 10% for drugs entering clinical trials, primarily due to a fundamental lack of understanding of underlying biology.
• 🧬 Progress is driven by two parallel revolutions: the AI revolution and a biology revolution enabling the creation of unparalleled amounts of human, cellular, and organismal data.
• 📈 Human genetics has already been shown to significantly increase the probability of clinical success by 2 to 3 times for drugs where it's deployed.

Insitro's Integrated Data & AI Approach

• 🔬 Insitro integrates multimodal human cohort data (genetics, quantitative pathophysiology measurements) with lab-generated human cell data (diverse genetics, omics, imaging).
• 🧠 AI amplifies data quality and scale by imputing unmeasured modalities, such as inferring liver fat from DEXA scans or blood markers, making large-scale cohorts more informative.
• ✅ AI helps define machine-learned disease axes, moving beyond binary classifications to understand the spectrum of healthy and sick states at both human and cellular levels.
• 🧪 The CRISPR-based "POSH" cell screening platform allows for high-throughput identification of genetic edits that shift cells from unhealthy to healthier states.

Accelerating Therapeutic Development

• 🎯 In fatty liver disease, AI-enabled discovery identified novel genetic modulators and highlighted de novo lipogenesis as a key pathway, leading to successful target validation in animal models within 18 months.
• ⚡ For ALS, Insitro identified a disease axis and screened for reverting treatments that address underlying pathologies like TDP43 mis-splicing, rapidly advancing towards a potential disease-modifying treatment.

Collaborative Innovation & Future Outlook

• 🤝 Partnerships are crucial for drug discovery, with collaborations spanning data providers (UK Biobank, Genomics England, NHS), drug developers (Bristol Myers Squibb, Lilly), and new initiatives.
• 👁️ A new partnership with Insights and Moorfields Hospital will develop an AI foundation model using NHS ophthalmic data (OCT scans) to uncover insights into various diseases, including cardiovascular, metabolic, and neurodegenerative conditions.
• 🌱 Insitro's mission is to bring better drugs faster to patients who can benefit most, leveraging machine learning and data at scale across metabolic, neurological, and ophthalmic diseases.