Genetically anchored drug target discovery for neglected diseases

Project Information

GenDrug

Grant agreement ID: 101116072

DOI

10.3030/101116072

EC signature date 13 February 2024

Start date 1 March 2024

End date 28 February 2029

Funded under

• European Research Council (ERC)

Total cost

€ 1 498 089,00

EU contribution

€ 1 498 089,00

Coordinated by

CHARITE - UNIVERSITAETSMEDIZIN BERLIN
Germany

Periodic Reporting for period 1 - GenDrug (Genetically anchored drug target discovery for neglected diseases)

Reporting period: 2024-03-01 to 2026-08-31

Summary of the context and overall objectives of the project

Most research funding is allocated to a comparatively small number of diseases commonly referred to as the major killers, such as cancers. However, hundreds of common non-communicable diseases substantially impair the quality of life of millions of individuals, even if they do not necessarily require hospital treatment. We refer to these as ‘neglected’ disease – not because they are rare, but because they are chronically underrepresented in research funding and, crucially, often lack safe and effective treatment.

GenDrug addresses this critical unmet need by developing innovative algorithms, including the responsible use of artificial intelligence, to guide the design of new drugs or identify opportunities to repurpose existing medications. Our core strategy exploits a powerful insight from human genetics: natural variation in our DNA sometimes mimics what a drug does to the body. When a genetic variant that reduces the activity of a specific protein also reduces disease risk, this provides compelling, human-relevant evidence that a drug targeting that same protein could be beneficial. Such genetic support is known to more than double the likelihood of a drug successfully reaching patients.

A concrete example: we were the first to show that variation near a gene called ADRA2A increases the risk of developing Raynaud's phenomenon, a painful condition affecting circulation in the fingers and toes (https://www.nature.com/articles/s41467-023-41876-5). This suggests that medications already used clinically to lower ADRA2A activity could be repurposed to treat this neglected condition. For drugs already in use, we will complement genetic evidence by emulating clinical trials using electronic health records from millions of patients, providing independent real-world support for repurposing candidates. Together, these approaches aim to deliver novel, evidence-backed treatment strategies for diseases that have long been overlooked.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

Human metabolism: We co-led the largest genetic study of blood metabolites to date, relating 249 molecules to genetic variation in nearly 450,000 people and identifying almost 30,000 associations. Variation altering the size of high-density lipoprotein particles influences coronary artery disease risk through a protein called VEGFA, pointing to a mechanism that acts alongside rather than through cholesterol lowering (Nature Genetics, 2025).

What shapes blood proteins: Modifiable characteristics such as medication, disease history and body composition explain on average more than twice as much variation in blood protein levels as genetic variation does. We established this across roughly 3,000 proteins in over 43,000 participants. MMP12 emerged as a causally supported marker for abdominal aortic aneurysm, a vascular condition for which no approved drug halts progression (Molecular Systems Biology, 2025).

Proteogenomics at scale: Pooling data from 78,664 participants across 38 studies, we identified more than 24,000 genetic variants influencing 1,116 blood proteins. The most influential regulators act not by changing how much protein is made, but by controlling how sugars are attached to it afterwards. Variants acting on proteins from a distance in the genome, largely overlooked until now, revealed repurposing opportunities including existing TYK2 inhibitors for rheumatoid arthritis (Cell, 2026).

Diverse populations: Mass-spectrometry proteomics in more than 1,400 British-Bangladeshi and British-Pakistani individuals, underrepresented in genomic research, yielded over 1,200 variant-protein associations, many specific to South Asian ancestries, and linked 21 proteins to 44 diseases (Nature Genetics, 2026).

Medical artificial intelligence: We interpreted disease signals that deep learning extracts from retinal photographs (Lancet Digital Health, 2026). A recent preprint tested the predictions of a model trained on the medical histories of hundreds of thousands of people against natural genetic variation, recovering established biology and pointing to treatable mechanisms behind the accumulation of disease over a lifetime (https://www.medrxiv.org/content/10.64898/2026.07.28.26359117v1).

Tools: We released the prodente software package and an interactive portal (https://omicscience.org/apps/prot_foundation/) so that researchers can test whether patterns in their own protein data reflect known health characteristics.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

Blood proteins for drug development: Variants that influence proteins from a distance in the genome reveal drug repurposing opportunities, a departure for a field that has relied almost exclusively on variants acting on a protein's own gene. These overlooked signals identify additional drug candidates and, in the same step, the measurable blood protein signatures that could select the patients most likely to benefit and show whether a treatment is working.

A target for an untreatable condition: MMP12 is a causally supported target for abdominal aortic aneurysm, a condition affecting around 5% of older men for which surveillance remains the only management. Three independent lines of evidence converge: variants raising blood MMP12 also raise aneurysm risk, MMP12 is elevated both in patients and in those who later develop the condition, and inhibiting MMP12 slows progression in animal models.

Interpretable medical AI: Models trained on patient histories and clinical images predict disease well but cannot account for their own predictions, which limits their clinical use. Testing those predictions against genetic variation offers a general route to extracting interpretable, potentially treatable biology from such models rather than accepting them as black boxes.

Equity across ancestries: Genomic research remains heavily biased towards European-ancestry populations, so genetically guided targets may not perform equally across patient groups. Our work in South Asian populations shows that clinically relevant signals are missed when only European cohorts are studied, and provides a template for other underrepresented groups. Realising this will require sustained investment in diverse biobanks, linkage to health records across healthcare systems, and engagement with clinical partners.