IR Sant Pau Coordinates European Project to Anticipate the Risk of Hemorrhage After Stroke

The Sant Pau Research Institute (IR Sant Pau) will coordinate BIOMERGE, an international project aiming to develop a tool capable of predicting which patients with ischemic stroke are at higher risk of suffering a severe cerebral hemorrhage after receiving thrombolytic treatment. The initiative will combine clinical information with genomic, epigenomic, and proteomic data using bioinformatics tools and artificial intelligence.

The project is led by Dr. Israel Fernández-Cadenas, head of the Pharmacogenomics and Neurovascular Genetics Group at IR Sant Pau, and brings together seven partners from six European countries. BIOMERGE has received total funding of €1,975,911. Of this amount, the Catalan Ministry of Health has awarded IR Sant Pau a grant of €249,720 to carry out the project between 2026 and 2028, under the transnational call of the European Partnership for Personalised Medicine (EP PerMed).

“We currently have highly effective treatments to restore blood flow during the acute phase of stroke, but it is still difficult to anticipate which patients may develop a severe hemorrhagic complication. BIOMERGE aims to generate tools that allow us to estimate this risk more accurately and move toward truly personalised management,” explains Dr. Fernández-Cadenas.

BIOMERGE was selected under EP PerMed’s JTC2025 call, which focuses on developing pharmacogenomic strategies to advance personalised medicine. The project aims to translate the analysis of large volumes of molecular information into a tool that can support decision-making during the acute phase of stroke, when the time available to start treatment is particularly limited.

A Serious Complication That Is Difficult to Anticipate

In ischemic stroke, a blood clot blocks blood flow to part of the brain. To dissolve the clot and restore circulation, thrombolytic treatments such as alteplase—also known as rtPA—or tenecteplase—TNK—may be administered, in addition to mechanical thrombectomy when indicated.

Although these treatments can decisively improve prognosis, some patients subsequently develop hemorrhagic transformation, or bleeding in the affected area of the brain. The most severe forms, known as PH-1 and PH-2 parenchymal hematomas, are associated with a higher risk of disability and death and may occur in approximately 5% to 10% of treated patients. Currently, there are no sufficiently accurate and rapid tools to anticipate which individuals are at greater risk of developing this complication.

Identifying these patients before treatment is administered could, in the future, make it possible to tailor certain clinical decisions, such as the intensity of blood pressure monitoring, the drug dose, or the choice of treatment. BIOMERGE is not yet starting from a tool that is ready for use in clinical care; rather, it seeks to generate and validate the biomarkers and algorithms needed to move toward that goal.

“This is not about limiting access to a treatment that can be decisive for a patient’s recovery but about having additional information to use it as safely and effectively as possible in each case,” the researcher notes.

Integrating Different Layers of Biological Information

The project will initially study a cohort of 450 patients with ischemic stroke treated with tPA or TNK. Half will have developed severe hemorrhagic transformation, while the other half will consist of patients with comparable characteristics who did not experience this complication.

Using blood samples collected before treatment, the researchers will analyze around 5,000 plasma proteins, approximately nine million genetic variants, and the methylation status of approximately 800,000 sites in the DNA. These data will be supplemented with multi-omics information already available from another 200 patients, enabling the team to work with an integrated cohort of 650 cases during the algorithm development phase.

Combining these different layers of information—genomic, epigenomic, and proteomic—will make it possible to identify biological signals that might go undetected if analyzed in isolation. The project will apply bioinformatics tools and machine-learning techniques to determine which combination of variables offers the greatest ability to predict individual risk.

The researchers will also use causal inference methods, such as Mendelian randomization, to distinguish biomarkers that are simply associated with hemorrhage from those that may be more directly related to the biological mechanisms causing it. Prioritizing the latter could improve the models’ ability to perform across different populations and facilitate their future clinical translation.

“One of the project’s main strengths is that we will not analyze genetics, epigenetics, or proteins independently. We want to integrate all this information to identify which combination of biomarkers provides a more robust prediction and, at the same time, better understand the mechanisms that explain treatment response,” says Dr. Fernández-Cadenas.

Two Predictive Models With Different Levels of Complexity

BIOMERGE will develop two complementary algorithms. The first will integrate clinical, genetic, epigenetic, and proteomic information to obtain the most comprehensive predictive model possible. The second will combine only clinical variables and plasma proteins, with the aim of producing a tool that is easier to bring into clinical practice.

This second approach addresses one of the main challenges in stroke treatment: decisions must be made in very little time. While certain genetic or epigenetic analyses still require complex procedures, some proteins can be detected in blood within minutes using rapid diagnostic technologies.

For this reason, the project will prioritize the identification of protein biomarkers that can be measured using devices in ambulances, even before the patient reaches the hospital. The long-term aim is for risk assessment to be integrated into the stroke care pathway and provide useful information during prehospital care.

Once developed, the algorithms will be validated in a new prospective cohort of 450 patients recruited from the stroke units of participating centers. This phase will make it possible to determine whether the models retain their predictive ability in patients different from those used to develop them. “For a tool of this kind to be useful, it is not enough for it to work in the cohort in which it was developed. We must demonstrate that it retains its predictive ability in new patients and can be integrated into the actual timelines and pathways of stroke care,” the researcher adds.

A European Consortium Coordinated From Sant Pau

The project’s scientific coordination is led by IR Sant Pau’s Pharmacogenomics and Neurovascular Genetics Group, headed by Dr. Israel Fernández-Cadenas. The team investigates the genetic, epigenetic, and molecular factors related to stroke, treatment response, and patients’ neurological outcomes.

In addition to IR Sant Pau, the consortium includes Leipzig University in Germany; Jagiellonian University in Kraków in Poland; the Institute of Biomedicine of Seville; University of Bordeaux in France; the Syreon Research Institute in Hungary; and St. Ann's University Hospital Brno in the Czech Republic. The consortium brings together experts in neurology, genetics, epigenetics, proteomics, bioinformatics, artificial intelligence, statistics, and health economics.

In addition to assessing the algorithms’ accuracy, BIOMERGE will analyze the feasibility of incorporating them into care pathways and their cost-benefit relationship compared with standard care. The project will also involve European patient and caregiver associations, which will help identify outcomes that are relevant from their perspective and assess the practical implications of the future tool.

“BIOMERGE’s international and multidisciplinary dimension is essential. We need to combine very different cohorts, technologies, and areas of expertise to ensure that the results are robust and can be applied beyond a single center or a specific population,” concludes Dr. Fernández-Cadenas.