Ep176: David Roblin on Building a TechBio Company

Episode Summary

Executive Summary: David Roblin traces Relation Therapeutics’ thesis: that AI works best when fused with rigorous wet-lab human biology, genetics, and clinical context to improve drug target discovery and reduce phase 2 attrition. He explains how his path from physician to pharma executive to translational leader shaped a patient-first, multidisciplinary approach in London’s Knowledge Quarter.

Main Topics: London’s Knowledge Quarter as a biotech hub (Priority: 5/5): Roblin describes the Knowledge Quarter as a dense cluster of universities, hospitals, institutes, and companies that concentrates intellectual capital and patient access for biotech R&D. Roblin’s journey from medicine to biopharma leadership (Priority: 5/5): He recounts a career that moved from Welsh upbringing and medical training into infectious disease, pharma development at Pfizer and Bayer, and translational research at the Crick. Why pharma fails and why human biology matters (Priority: 5/5): Roblin argues that phase 2 attrition remains the central bottleneck, and that too much drug discovery still relies on top-down hypotheses and non-human models instead of human tissue and genetics. The founding thesis of Relation Therapeutics (Priority: 5/5): Relation combines multi-omics, machine learning, and clinical samples to identify disease-driving relationships and better targets, while emphasizing collaboration across scientific disciplines. Platform strategy: genetics, single-cell biology, and lab-in-the-loop learning (Priority: 5/5): He outlines Relation’s core workflow, including its DNA foundation model, proprietary tissue methods, CRISPR perturbation assays, and feedback loop that refines target selection. Lead programs and GSK partnership (Priority: 4/5): Osteoporosis is Relation’s proprietary lead program, while collaborations with GSK extend the platform into fibrosis and osteoarthritis where large-pharma expertise and shared data can accelerate discovery. What AI can and cannot do in drug discovery (Priority: 4/5): Roblin says AI is a tool within a broader integrated system, not a replacement for wet biology, and warns against overselling 'AI-discovered drugs' as distinct from conventional biology-driven discovery.

Key Arguments: Human disease biology is still poorly understood, and the biggest opportunity is to improve target selection by studying human tissue, not just animal models. Phase 2 remains the key attrition point in drug development, so better biology and better target discovery are needed upstream. Machine learning becomes most useful when it is integrated with wet lab experiments, clinical samples, and R&D judgment rather than used in isolation. Relation’s value lies in connecting different data types—genetics, single-cell data, transcriptomics, proteomics, and phenotypes—to uncover causal relationships. Interdisciplinary teams work best when scientists understand each other’s language and question framing; culture is part of the platform. Osteoporosis was chosen because it has strong genetic signal, clear unmet need, measurable translational biomarkers, and workable single-cell biology. A successful AI model should rediscover known biology first; if it does that reliably, it can then suggest novel targets worth testing. Big-pharma partnerships make sense in areas like osteoarthritis and fibrosis where Relation’s platform can complement deep disease expertise and development capabilities.

Data Points: Funds raised to date: more than $80 million - Relation’s cumulative financing from investors with tech and biotech backgrounds GSK collaboration timing: late last year - Preclinical partnership announced for fibrosis and osteoarthritis discovery work Knowledge Quarter scale: roughly 1 mile square - Area around King’s Cross where Relation is based Patients accessible through local hospitals: 20 million patients - Roblin cites the hospital network around the Knowledge Quarter as a major asset Crick Institute size: over 1 million square foot - Integrated wet-lab facility central to the Knowledge Quarter ecosystem Pfizer-Sandwich historical footprint: about a third of Pfizer’s research efforts in the UK - Roblin describes the former scale of Pfizer’s UK research presence Phase 2 attrition: 80% to 90% - Roblin says attrition remains high across therapy areas CCR5 heterozygous rate: 20% - Roblin explains the human genetics behind HIV resistance research CCR5 homozygous rate: 1%–2% - People with lack of function in CCR5 are strongly protected from HIV entry Maraviroc approval year: 2007 - Example of a drug program motivated by human genetics Crick group-leader tenure model: 6 plus 6 years - Junior group leaders are reviewed after six years and can extend another six Crick group size cap: 10 to 12 people - Small teams were designed to encourage collaboration Initial Relation seed funding: just over $1 million - From Juvenessence for the company’s start Additional early grant funding: $1.6 million - Bill & Melinda Gates Foundation support for combinatorial/drug-repurposing work Current team size: 70 people - Relation’s headcount at the time of the interview Bone samples analyzed: about 300 - Relation’s osteoporosis work enabled a large bone atlas Human genome genes: 22,000 genes - Roblin describes the scale of the genome used in target discovery Genes expressed per cell: around 8,000 genes - General explanation of cell-specific transcriptional activity Disease variants in osteoporosis: over 1,100 DNA variants - GWAS signal supporting osteoporosis as a platform-fit indication Osteoporotic fracture risk: 1 in 4 for men; 1 in 2 for women - Roblin highlights the major unmet need in aging populations Increase in search space: 143% - Relation’s model expanded the set of candidate osteoporosis genes

Pivotal Quotes: "I like to be the most stupid person in the room." — David Roblin: He explains his leadership style: hiring experts and staying comfortable not being the technical authority in every domain "The patient is waiting." — David Roblin: He states Relation’s mission as a patient-centered reason for doing the work "We are drowning in a sea of data but remain hungry for knowledge." — David Roblin (quoting Sidney Brenner): Used to describe why new data-rich technologies need better integration and interpretation

Implications: Relation’s approach suggests AI drug discovery will succeed only when anchored to human biology, not hype. The model could become a template for future biotech: multidisciplinary, sample-driven, and tightly coupled to clinical reality.