Scientists have created an AI-driven “map” of the heart that links medical images with genes and drugs, revealing fresh leads for treating heart disease and repurposing existing medicines.
A new artificial intelligence tool that fuses heart scans with vast troves of biological data could help researchers find heart disease drugs faster — including new uses for medicines already on the market.
The system, called CardioKG, is the first knowledge graph to weave detailed images of the heart’s structure and function into a large, searchable network of genes, diseases, drugs, symptoms and molecular pathways. It was developed by a team led by postdoctoral researcher Khaled Rjoob and group leader Declan O’Regan of the Computational Cardiac Imaging Group at the MRC Laboratory of Medical Sciences.
Knowledge graphs are a way of organizing information as a web of connected facts rather than as isolated entries in separate databases. In biomedicine, they can link what is known about genes, proteins, diseases and treatments, helping scientists spot patterns and relationships that might otherwise be missed.
Until now, though, these graphs have mostly lacked one crucial layer: how organs actually look and behave inside the body.
To change that, the CardioKG team turned to the UK Biobank, a long-running health study that includes imaging and genetic data from hundreds of thousands of volunteers. The researchers pulled heart imaging data from 4,280 participants with atrial fibrillation, heart failure or heart attack, along with 5,304 healthy participants. That allowed them to capture a wide range of variation in how hearts are shaped and how well they pump.
From these scans, the team generated more than 200,000 image-based traits — precise measurements that describe features such as chamber size, wall thickness and how different parts of the heart move during each beat. They then integrated those traits with information from 18 different biological databases and used AI to learn how all of these pieces fit together.
The result is a richly detailed map that can be used to predict which genes are linked to specific heart problems and which existing drugs might help treat them.
The power of the approach comes from its ability to connect many types of information in one place, according to O’Regan, who is also a professor of cardiovascular AI at Imperial College London.
“One of the advantages of knowledge graphs is that they integrate information about genes, drugs and diseases,” O’Regan said in a news release. When imaging data were added, “this means you have more power to make discoveries about new therapies. We found that including heart imaging in the graph transformed how well new genes and drugs could be identified.”
Using CardioKG, the model flagged a list of previously unrecognized genes associated with heart disease. It also pointed to two existing drugs that might be repurposed for heart conditions.
One, methotrexate, is commonly used to treat rheumatoid arthritis. The graph’s predictions suggest it could improve outcomes in heart failure. Another class of drugs, gliptins, currently prescribed to manage diabetes, emerged as potential candidates to benefit people with atrial fibrillation, a common heart rhythm disorder.
In a surprising twist, the analysis also suggested that caffeine, which is known to make the heart more excitable, may have a protective effect in patients with atrial fibrillation who have an irregular and fast pulse. That finding aligns with a growing body of research suggesting that, for some people, moderate caffeine intake may not be as harmful to heart rhythm as once feared, though more work is needed before it could influence clinical advice.
O’Regan noted that the team’s early results are not appearing in isolation.
“What’s exciting is there are other recent studies in the field which support our preliminary findings,” he said. Taken together, “this highlights the huge potential of knowledge graphs in uncovering existing drugs that might be repurposed as new treatments.”
Drug repurposing is especially attractive in cardiology because it can dramatically shorten the time and cost needed to bring new therapies to patients. Medicines that are already approved have known safety profiles, so if a tool like CardioKG can reliably point to new uses, researchers can move more quickly into targeted clinical trials.
Beyond specific drug leads, the technology offers a new way to prioritize biological targets. By rapidly generating ranked lists of genes likely to be involved in particular heart conditions, CardioKG could give pharmaceutical companies and academic labs a more efficient starting point for discovery. Instead of sifting through thousands of possibilities, scientists can focus on the most promising candidates highlighted by the graph and then validate them in experiments.
The researchers see CardioKG as a proof of concept that could extend far beyond the heart. Similar knowledge graphs could be built for any organ that can be imaged, such as the brain or liver, or for whole-body scans that track body fat and other tissues. That could open new avenues for tackling conditions like dementia, obesity and metabolic disease by linking how organs look on scans to the underlying biology and potential treatments.
The next step is to make the graph more personal and dynamic, so that it reflects how disease unfolds over time in real people, according to Rjoob.
“Building on this work, we will extend the knowledge graph into a dynamic, patient-centred framework that captures real disease trajectories,” he said in the news release. “This will open new possibilities for personalised treatment and predicting when diseases are likely to develop.”
The study, published in the journal Nature, was supported by the Medical Research Council, the British Heart Foundation, Bayer AG and the National Institute for Health and Care Research Imperial College Biomedical Research Centre.
As heart disease remains a leading cause of death worldwide, tools that can connect the dots between images, genes and drugs may help turn the tide — not by inventing entirely new medicines from scratch, but by revealing new ways to use the ones we already have, and by pointing the way to smarter, more personalized care.
Source: Medical Research Council (MRC) Laboratory of Medical Sciences