Researchers from the Sant Pau Research Institute (IR Sant Pau) and CIBERER have developed an AI tool that identifies hundreds of molecular signals to predict venous thrombosis. By integrating clinical, genetic, and transcriptomic data, the system improves risk stratification for idiopathic venous thromboembolism, as detailed in the Journal of Thrombosis and Haemostasis.
Why do some people develop thrombosis without clear risk factors?
Traditional medicine focuses on visible triggers like obesity, age, or hormone treatments. However, many patients suffer from idiopathic venous thromboembolism—cases where no clear cause exists. According to the IR Sant Pau study, genetic factors influence more than 60% of the individual variability in thrombosis risk, yet known hereditary markers don’t explain every case.
To close this gap, Dr. Pol Ezquerra and his team analyzed 790 people from the GAIT2 (Genetic Analysis of Idiopathic Thrombophilia) family cohort. This included 70 individuals who had previously experienced idiopathic venous thrombosis. They looked beyond simple DNA sequences, examining the activity of 12,981 genes to see how they actually behave in the body.
How does AI improve the accuracy of thrombosis risk profiles?
The researchers used machine-learning algorithms to process thousands of biological variables simultaneously. This approach identified 494 genes whose activity distinguishes those who have had thrombosis from those who haven’t. While the AI still recognized standard markers—like body mass index, age, and von Willebrand factor levels—it added a layer of “molecular signatures” that traditional tests miss.

The impact on accuracy is concrete. According to Dr. José Manuel Soria, director of the Complex Disease Genomics Unit at IR Sant Pau, integrating these variables allows for a more accurate description of risk profiles than analyzing factors in isolation.
| Model Type | False High-Risk Rate (No History) | Detection Rate (History of Disease) |
|---|---|---|
| Clinical & Genetic Only | 43% | 70% |
| With Transcriptomic Data | 23% | 74% |
What happens next for personalized prevention?
The tool creates a “similarity score” that measures how closely a person’s molecular profile matches those who have already suffered a thrombotic event. This means doctors could potentially identify high-risk individuals before a clot ever forms. The study also found links to cardiovascular and renal processes, specifically molecular pathways related to cardiomyopathies and the kidney’s proximal tubules.
While the tool requires validation in independent cohorts before clinical use, Dr. Soria notes that these strategies could eventually lead to preventive measures tailored to each specific patient’s molecular activity.
Frequently Asked Questions
What is idiopathic venous thromboembolism?
It is a form of venous thrombosis that occurs without any clear triggering factors or known clinical risks.

How does this AI tool differ from a standard blood test?
Standard tests look for specific markers or clinical signs. This AI tool integrates transcriptomic data—the activity of nearly 13,000 genes—to find patterns that a human doctor or a single-marker test would overlook.
Can I use this tool for diagnosis today?
No. The researchers emphasize that the tool still needs validation in independent groups before it can be applied in a clinical setting.
Want to stay updated on the future of precision medicine? Subscribe to our newsletter or leave a comment below sharing your thoughts on AI-driven diagnostics.














