The UKB-PPP consortium should receive the first tranche of ~100k samples this autumn and subsequent tranches throughout 2026. Data releases to all approved UKB researchers will begin next year.

*Alden Scientific, Amgen/deCODE genetics, AstraZeneca, Bristol Myers Squibb, Calico Life Sciences, Genentech, GSK, Isomorphic Labs, Johnson & Johnson, Merck Group, Novo Nordisk, Pfizer, Regeneron and Takeda.

I'm deeply grateful to the hundreds of collaborators working across fourteen biopharmaceutical companies*, two technology vendors, and one remarkable biobank for making this possible.

This initiative again underpins the transformative value of public-private partnerships for drug development and healthcare and reemphasizes the quiet 'revolution' happening in the field of proteomics.

This will allow us to build foundational AI models for biological discovery, study biomarkers that change over time in response to changes in health states, identify surrogate blood biomarkers for MRI endpoints, and - ultimately - develop better medicines & diagnostics.

We will begin by profiling 300k samples, comprising approximately 250k samples collected at baseline visits and 50k samples taken at various follow-up assessments. We aim to incorporate an additional 250k baseline samples and 50k repeat samples pending additional funding.

Could be combined in the future with a continuous sensor (implant) for real-time detection of proteins driving inflammation in high-risk individuals
www.science.org/doi/10.1126/...
@science.org

Possibly, but Alamar is low-plex by design (n~250) & captures many low-abundance proteins - not necessarily the most widely studied. Mike’s comment & Jochen’s review offer important perspectives; abundance levels seem to influence confidence in cross-platform detection (higher = more confidence).

Here we tried to provide some objective views to the reoccurring discussion:

pubs.acs.org/doi/10.1021/...

The key is to understand the biases and factors causing the differences in the data. Those can include and do go beyond the antibodies.

Yes, “good” is relative to what’s been reported before. It’s the median correlation across 217 commonly measured proteins, with a wide range overall. More documentation of the antibodies used +/ epitopes bound could partially address the problem, but might be impractical to implement.

We’ve found Olink & Alamar to correlate quite well (~60%) overall. Maybe unsurprising, since both are antibody-based.

Lack of correlation b/w SomaScan & others might reveal interesting facets of the aptamer tech (protein folding, PTMs, etc) but 100% agree that infighting is bad for the field.