Heart failure (HF) is a severe cardiovascular condition, with alarming mortality. The challenge of identifying patients at significant risk for HF with preserved ejection fraction (HFpEF) is critical. Treatment such as mineralocorticoid receptor antagonists (MRA) offer hope but come with notable side effects. The glycocalyx of endothelial cells (ECs) participates in the proper maintenance of vascular function by allowing ECs not to be in direct contact with the blood. Our previous results highlight a central question: is the degradation of the EC glycocalyx prior to the development of endothelial dysfunction and the development of cardiovascular pathologies? The integration of omics and machine learning (ML) has emerged as a potent combination to advance our understanding of complex interactions. Such combination is of interest in complex mechanism regulation as the changes in EC glycocalyx in HFpEF (diagnosis and treatment). The insights gathered from the glycocalyx exploration in the complexities of HF, augmented by the precision and depth of omics and ML, are capable of deciphering the complex pathophysiology underlying HFpEF genesis but also the response to preventive treatment. This project will (1) evaluate the implication of glycocalyx degradation as the early step of HFpEF genesis and (2) during HFpEF treatment. (3) Identify the underlying/connecting integrated pathophysiology surrounding glycocalyx alterations in HFpEF. (4) Provide an integrated framework to identify, with a software, patients most likely to a) develop HFpEF and b) benefit from MRA as a preventive HFpEF treatment. Cohorts and clinical trials of general population, patients developing or with established HFpEF and receiving MRA will be used to understand the time course of glycocalyx degradation. With proteomic results and ML we will identify biological framework of HFpEF development and response to MRA that will serve to create a software dedicated to help therapeutical decision.

The BCI4IA project aims to design a brain-computer interface for detecting intraoperative awareness during general anesthesia (GA). No satisfactory solution exists when this phenomenon causes severe post-traumatic stress disorder. The first reaction of patients is to move, but the curarization used in surgery paralyzes them. We want to study motor brain activity under GA using electroencephalography to detect markers of motor intention. We will observe a combination of markers (relative powers, connectivity ...) under propofol with and without a median nerve stimulation to amplify them and we will design non-existent classification methods by adapting transfer learning in Riemannian geometry for this problem where only the examples of one class are available, that of rest, since we have no examples of motor intention of the patient under GA.