Biology is data.
AI reads it.
Ageing is not a single event — it is a cascade of molecular decisions made by billions of cells over decades. We build the intelligence to decode those decisions, predict their trajectories, and ultimately reverse them.
How We Research
A four-stage pipeline from raw biological data to validated longevity interventions — built for rigour and designed for speed.
Multi-Omic Data Ingestion
We ingest genomic, proteomic, metabolomic, and epigenomic datasets from diverse cohorts — harmonising measurement modalities and resolving batch effects before a single model sees a single row.
Biological Age Modelling
Using methylation arrays, transcriptomic signatures, and composite biomarker panels, we train biological age clocks that outperform chronological age in predicting all-cause mortality and disease onset.
Pathway-Level AI Inference
Large graph neural networks learn the topology of FOXO-regulated pathways — surfacing which molecular switches are most predictive of resilient ageing and which are most amenable to intervention.
Intervention Prioritisation
Candidate interventions — dietary, pharmacological, lifestyle — are ranked by predicted effect size on biological age trajectories, cross-validated against longitudinal outcome data before advancing to experimental validation.
What Our Models Learn From
Our AI is only as good as the data it learns from. We are rigorous about source quality, temporal depth, and outcome linkage.
Longitudinal Cohort Data
Multi-year follow-up datasets tracking thousands of biomarkers per individual — giving us the temporal resolution to model ageing trajectories rather than snapshots.
Methylation Arrays
Genome-wide DNA methylation profiling across age-diverse populations — the substrate for next-generation epigenetic clock construction and validation.
Proteomic Panels
High-plex plasma proteomics covering thousands of circulating proteins — capturing systemic biological age signals that single-biomarker panels miss entirely.
Clinical Outcomes
Linkage to hospital records, mortality registries, and disease incidence data — grounding our models in hard clinical endpoints, not surrogate proxies.
Intervention Trials
Randomised and quasi-experimental data from dietary, pharmacological, and lifestyle interventions — training causal models, not just correlational ones.
Wearable & Continuous
Continuous physiological monitoring data — HRV, glucose, sleep architecture, activity — filling the gaps between clinic visits with real-world biological signal.
Where AI Meets Biology
The principles that guide every modelling decision, every dataset choice, and every claim we make about what our AI can and cannot do.
Causal over correlational
Correlation tells you what tends to happen. Causality tells you what to do about it. We design our models and experiments to distinguish the two — because intervening on a correlation that is not causal achieves nothing.
Multi-omic by default
No single data layer explains ageing. Genomics without proteomics misses post-translational reality. Epigenetics without metabolomics misses energy state. We integrate all layers or we don't model at all.
Populations and individuals
Population-level findings drive discovery. Individual-level models drive intervention. We build both — because the biology that predicts average mortality does not predict your mortality.
Open to being wrong
Longevity biology is young. Our models will be wrong in ways we cannot yet see. We ship early, validate rigorously, update aggressively, and never mistake confidence for correctness.