Knowledge graphs are a strong tool for combining information from biological databases and connecting existing knowledge about genes, diseases, treatments, molecular pathways, and symptoms in a structured network. Previously, they have lacked detailed, individual-level information about how the affected organ actually appears and operates.
Recent research has enhanced this technology by integrating imaging data into a knowledge graph for the first time.
CardioKG provides a comprehensive view of the heart's structure and function. It could be used to enhance the precision of predicting gene-disease connections and evaluating the potential of existing drugs for treatment.
Capturing Heart Variation
To create CardioKG, the team used heart-imaging data from 4,280 UK Biobank participants diagnosed with atrial fibrillation, heart failure, or heart attack, along with 5,304 healthy participants, documenting variations in heart structure and function.
Over 200,000 image-based traits were produced and employed to train the model. The team combined this data with information from 18 different biological databases, using artificial intelligence (AI) to forecast gene-disease connections and possibilities for drug repurposing.
One of the advantages of knowledge graphs is that they integrate information about genes, drugs and diseases, 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.
Declan O’Regan, Professor, Computational Cardiac Imaging Group, MRC Laboratory of Medical Sciences
Predicting New Drug Opportunities
The model identified a list of new genes associated with diseases and predicted two potential drugs for treating heart ailments. Methotrexate, a medication commonly used for rheumatoid arthritis, may improve heart failure; Gliptins, used in diabetes treatment, could benefit patients with atrial fibrillation.
The team also surprisingly found that caffeine, which increases heart excitability, has a protective effect in atrial fibrillation patients with an irregular and rapid heart rate.
What’s exciting is there are other recent studies in the field which support our preliminary findings, this highlights the huge potential of knowledge graphs in uncovering existing drugs that might be repurposed as new treatments.
Declan O’Regan, Professor, Computational Cardiac Imaging Group, MRC Laboratory of Medical Sciences
Extending the Technology to Other Organs
CardioKG is a proof-of-concept technology, and its applications could extend far beyond the heart. Researchers could create knowledge graphs integrating imaging data for any organ, including brain scans, body-fat imaging, and other tissues, to explore new therapeutic options for conditions like dementia or obesity.
These knowledge graphs can quickly and accurately generate lists of high-priority genes for various diseases. This would provide pharmaceutical companies with a valuable starting point by identifying biological targets for exploration, validation, and potential development into new therapies more efficiently than traditional methods.
Building on this work, we will extend the knowledge graph into a dynamic, patient-centered framework that captures real disease trajectories. This will open new possibilities for personalized treatment and predicting when diseases are likely to develop.
Dr. Khaled Rjoob, Computational Cardiac Imaging Group, MRC Laboratory of Medical Sciences