Thunder

Mon, 01 Sep 2025 00:00:00 +0000

Machine learning (ML) offers transformative opportunities for healthcare, with applications ranging from precision medicine to operational optimization. However, progress is constrained by limited access to diverse, high-quality datasets, exacerbated by fragmentation, data scarcity, and stringent privacy regulations. Traditional data augmentation methods fail to fully capture the complexity and heterogeneity of healthcare data. Generative AI, particularly large language models (LLMs), offers a promising alternative by synthesizing realistic datasets while addressing data scarcity. Yet, their adoption in healthcare is hindered by critical concerns about trustworthiness, including semantic validity, fairness, bias mitigation, fidelity, privacy preservation, and real-world utility. This research identifies key gaps in developing trustworthy generative models and ML methods for healthcare. These include the absence of standardized synthetic data evaluation frameworks, trust deficits in healthcare generative models, resource intensiveness, and design-induced opaqueness. The overall objective of the THUNDER project is to forge a comprehensive framework for trustworthy and responsible generative AI in healthcare. This will be achieved by defining (i) standardized evaluation metrics, (ii) developing advanced knowledge-guided generative models, and (iii) creating a fully frugal-by-design and interpretable-by-design learning models. These efforts, driven by interdisciplinary and intersectoral mobility and knowledge exchange, will establish a new paradigm for AI-driven healthcare. We will target Sepsis as a global health priority identified by the World Health Organization (WHO).

Systemic Sclerosis

Tue, 27 Sep 2022 00:00:00 +0000

Systemic sclerosis is a severe and complex disease

Systemic sclerosis (SSc) is a rheumatological disease belonging to the group of connective tissue diseases. These autoimmune diseases are rare but severely impact the quality of life of patients. SSc is the one associated with the highest morbimortality. It usually occurs in women in the 5th decade. The pathophysiological mechanisms that cause SSc are not fully elucidated. As the name suggests, it is characterized by fibrosing skin involvement, i.e., cutaneous soft tissues are progressively replaced by collagen deposits that constitute an inert material responsible for the thickening of the skin that becomes difficult to pinch. Genetic factors are probably involved, as is often the case with autoimmune diseases (immune predisposition genes). There are probably environmental factors, like those recognized in occupational forms of the disease, often in men exposed to silica for example.

SSc is associated with diverse manifestations. The majority of patients present a Raynaud’s phenomenon (painful discoloration of the fingers at temperature changes) which is often severe and can be complicated by digital ulcerations (painful wounds, necrosis of the fingers). This is characteristic of vascular involvement which can reach other small vessels such as the pulmonary circulation, resulting in a very severe complication named pulmonary hypertension. Almost any organ or system can be affected by the inflammation or fibrosis associated with the disease: pulmonary fibrosis responsible for coughing, shortness of breath or even lack of oxygen, cardiac damage responsible for intolerance to effort, heart rhythm disorders with syncope and sudden death, digestive damage with gastroesophageal reflux (very annoying burns in daily life), bloating, severe constipation with anal incontinence or, on the contrary, significant diarrhoea with malnutrition, musculoskeletal damage with pain, joint stiffness that can be fixed. The aesthetic consequences are marked. These damages are responsible for major personal, family and socio-professional burden. They have a strong impact on patient survival, as shown by a study of the French cohort, where we found a survival rate of 72% at 10 years, and a standardized mortality ratio of 3.45 (95%CI 3.03-3.94) in the metaanalysis of worldwide studies.

Classification of SSc into subsets needs to evolve

The management of this condition is complex, therefore stakeholders are increasingly seeking to establish recommendations. Early diagnosis is important as dedicated treatment may slow disease progression and avoid complications. However, there is often a significant delay in diagnosis due to a lack of knowledge of the disease or the complexity of the presentations, which slows down the referral of the patient to an expert. Once diagnosed, the clinician in charge of the patient will try to predict the severity of the disease and in particular the potential organ damage. To do this, we apply a classification of the disease into subgroups that has been widely used for more than 30 years. This classification is based on the cutaneous extension of the fibrosis and defines two distinct groups of patients: the limited cutaneous form (lcSSc), where the cutaneous fibrosis does not extend beyond the elbows and knees, and the diffuse cutaneous form (dcSSc), where the cutaneous fibrosis also affects the thorax, the abdomen or the proximal parts of the limbs. Although this classification has had important value in distinguishing groups of patients with different prognoses and was based on a relatively simple clinical assessment, it is now recognized that it is insufficient and does not always correspond to what is seen in clinical practice. The main problem is that it does not reflect the heterogeneity of the disease, especially in terms of organ involvement. For example, we can see patients with dcSSc who are doing relatively well without major organ damage, and others with lcSSc who will die from pulmonary fibrosis. The other issue raised by this classification are the impact on:

(i) patients (e.g. those with dcSSc who will fear complications that will not always occur) and
(ii) research (efforts have been focused on dcSSc to the detriment of lcSSc, even though we now know that this form is also associated with a major impact on quality of life).

It demonstrates that the classification of a disease has a very important impact on its management and medico-economic policies. Numerous studies have suggested that SSc is in fact a much more heterogeneous disease than the simple dichotomy between lcSSc and dcSSc with diverse clinical presentations and varying degrees of organ involvement among patients. We conducted an unsupervised learning analysis of the European Scleroderma Trials and Research (EUSTAR) cohort comprising 137 centers i.e. ~7000 patients. Cluster analysis confirmed that there was significant heterogeneity and that organ involvement and antibody profiles played an important role in cluster discrimination. We showed that some clusters resembled the classically expected forms (older women with lcSSc and mild organ involvement, younger men with very severe dcSSc) but also highlighted subsets less described in the literature but corresponding to clinical practice.

Translational research: understanding the pathogenicity of antinuclear antibodies in systemic sclerosis.

Antinuclear antibodies (ANA) are present in the vast majority of patients with systemic sclerosis (SSc). ANA are valuable tools in the management of patients with SSc. Indeed, they represent the best biomarkers of the disease, as they are strongly associated with distinct clinical phenotypes and prognoses. For example, anti-topoisomerase I antibodies (ATA) are associated with more severe forms of the disease and with the presence of pulmonary involvement. However, their role in the pathophysiology of the disease has not been demonstrated.

Within the ENDOMIC team, we have shown through omics studies that immunoglobulin G (IgG) from patients with SSc modifies the molecular profile of normal fibroblasts and endothelial cells, depending on the ANA serotype. In particular, we identified distinct molecular profiles induced by ANA-positive IgG.

Based on these findings, we hypothesize that ANA from SSc patients directly interact with their intranuclear targets. Using combined approaches including confocal microscopy, super-resolution microscopy, and live-cell imaging, we demonstrated that ATA can access the nucleus of living cells and interacting with their nuclear target antigen (topoisomerase I). These data suggest a paradigm shift in systemic autoimmune diseases, in which ANA would no longer be considered merely spectators and biomarkers, but also active players in pathophysiological mechanisms.

machine learning on ENDOMIC Team

Thunder

Systemic Sclerosis

Systemic sclerosis is a severe and complex disease

Classification of SSc into subsets needs to evolve

Translational research: understanding the pathogenicity of antinuclear antibodies in systemic sclerosis.