Chemistry in Healthcare: From Phenomics to AI Driven Medicines

Podcast at a glance

The episode surveys how chemistry underpins modern medicine, from genomics limitations to phonomics, scalable biomarker tests, and new computational approaches that could design medicines from first principles. It also highlights a menstrual health biobank project and a conversation about how AI and deep learning are accelerating drug discovery and therapies such as CAR T cells and MRNA vaccines.

• Key shift from genetics to environment and chemistry driven phenotyping
• Menstrual health biobank aiming to identify iron deficiency risk from menstrual fluid
• Exploration of how chemistry enables scalable cardiovascular risk tests
• AI and deep learning enabling generative design of biomolecules

Introduction: Chemistry as a pillar of modern medicine

The podcast opens by framing chemistry as foundational to discovery, diagnostics and treatment in health care. It stresses that while reading the DNA code reveals potential disease risks, the actual manifestation depends on genome environment interactions with diet, lifestyle and physiology. In this context, phonomics or metabolomics emerges as a central approach to measure thousands to millions of molecules in the body to correlate them with disease outcomes. Jeremy Nicholson from the Hong Kong University of Science and Technology is highlighted as a pioneer, discussing how chemical analyses of body fluids can inform disease risk and treatment decisions.

Genomics vs Phenomics: Limits and opportunities

The conversation emphasizes the bottlenecks of relying solely on genomics for population screening. Though genetics can identify some high risk markers in targeted areas like BRCA or BRCA related cancers, whole genome sequencing remains impractical at scale. The environment, lifestyle, and interactions among multiple genes are not captured by genome data, limiting its predictive power for diseases like cardiovascular disease. The podcast argues for scalable, translatable technologies that capture metabolic signatures in blood and other biofluids as an alternative route to health risk assessment and disease management.

From discovery to scalable testing: biomarker markers and technology

The discussion covers the shift from discovering markers to building practical tests. The aim is to identify markers that predict major diseases, then simplify the tests so they can be deployed outside large hospitals. Advances in mass spectrometry and NMR are highlighted, including newer instruments capable of analyzing hundreds of thousands to a million samples rapidly. The role of computation and AI is stressed to manage the data deluge and to help translate discoveries into usable diagnostics at scale.

Menstrual health biobanking: Miss Vital Sign

The transcript features Gemma Sharp describing a world’s largest menstrual fluid biobank as part of the Miss Vital sign program. The study follows women over three menstrual cycles to collect comprehensive phenotypic data and dried menstrual fluid on filter papers inside a specialized pad, enabling remote collection and easy transport. The cohort includes women aged mid-30s to mid-40s from Bristol and Bradford, focusing on heavy menstrual bleeding and iron deficiency risk. The study uses questionnaires as the main data collection tool and compares questionnaire definitions with an alkaline hematin gold standard that is impractical for large scale use in clinics, seeking a middle ground between objectivity and practicality.

Chemistry in diagnostics and drug discovery

The podcast shifts to how chemistry underpins therapeutic strategies, including the mechanisms of common analgesics like paracetamol and aspirin, and links to historical drug development such as morphine. It outlines the central pharmacological targets—enzymes and receptors—and notes how modern discovery leverages bench assays and high-throughput screening to explore millions of compounds. A key turning point is the move from discovery by trial and error toward rational target identification and assay-based screening, enabling scalable drug development and the potential for patentable therapeutics.

The AI frontier in medicine: design from first principles

David Baker discusses how deep learning empowers the design of new biomolecules that fit receptor structures, a process likened to crafting a key for a lock. Generative AI can propose novel protein sequences that bind targets, which are then validated in cells and model disease systems. While proteins present the easiest starting point because they can be encoded genetically and produced in the lab, extending these design principles to other molecules remains more challenging due to data scarcity and synthesis hurdles. The conversation also touches on applications beyond conventional drugs, including enhancing CAR T cells and integrating protein design with mRNA vaccines to deliver functional proteins.

CAR T therapies, RNA vaccines and the broader future

The discussion highlights how design proteins can augment cellular therapies such as CAR T cells, and how mRNA platforms can deliver designed proteins as vaccines or therapeutics. The potential for AI to contribute beyond small molecule design—into biologics and gene therapies—reflects a broader trend toward AI-assisted, precision interventions in oncology, immunology and metabolic diseases.

Conclusion: toward a trusted, AI-augmented future of health

The podcast closes with reflections on the necessity of combining chemical knowledge, scalable technologies, and computational power to realize a future where health care is more personalized, accessible, and grounded in validated science. The program is presented by Naked Scientists with support from Leilani Aro Smith and a note of thanks to contributors, inviting support from listeners via donations.