HAL UVSQ
Not a member yet
57319 research outputs found
Sort by
sCellST predicts single-cell gene expression from H& E images
No full textInternational audienceAbstract Understanding the spatial organization of individual cell types within tissue and how this organization is disrupted in disease, is a central question in biology and medicine. Hematoxylin and eosin-stained slides are widely available and provide detailed morphological context, while spatial gene expression profiling offers complementary molecular insights, though it remains costly and limited in accessibility. Predicting gene expression directly from histological images is therefore an attractive goal. However, existing approaches typically rely on small image patches, limiting resolution and the ability to capture fine-grained morphological variation. Here, we introduce a deep learning approach that predicts single-cell gene expression from morphology, matching patch-based methods on spot level prediction tasks. The model recovers biologically meaningful expression patterns across two cancer datasets and distinguishes fine cell populations. This approach enables molecular-level interpretation of standard histological slides at scale, offering new opportunities to study tissue organization and cellular diversity in health and disease
Santé mentale en tension : enjeux épistémologiques, politiques et cliniques d'un espace fragmenté
No full textInternational audienc
Real-life use of bone-targeting agents for bone metastases in France between 2009 and 2018: Results of the OPTIMOS study
No full textInternational audienc
Subacute loss of olfactory neurons following SARS-CoV-2 infection in hamsters
No full textInternational audienceThe loss of smell has been a hallmark of the COVID-19 pandemic. Odorant detection relies on neurons present in the olfactory epithelium supported by sustentacular cells. The latter are massively infected by SARS-CoV-2 along with infiltration of innate immune cells and desquamation of the olfactory epithelium. This destruction leads to release of olfactory epithelium cells into the lumen of the nasal cavity, but the extent of the loss of mature olfactory neurons remains to be clarified. In this study, we compared the spatiotemporal evolution of the olfactory epithelium during SARS-CoV-2 infection with associated smell impairment in hamsters. The olfactory performance of infected hamsters decreases as early as 2 days post-infection (dpi), then gradually recovers through 17 dpi. While the infection is mostly resolved after 4 dpi in the nasal cavity, we observed a subacute decrease of the mature olfactory neuron population which almost completely disappear at 11 dpi. Furthermore, regeneration of the olfactory epithelium does not start until 8 dpi and leads to a high fraction of immature olfactory neurons. The delayed regeneration and persistent alteration of the olfactory neuron population was correlated with a prolonged expression of inflammatory cytokines and a rapid decrease of the levels of anti-inflammatory markers linked to regeneration. Overall, our results suggest that the regeneration process is altered in some areas of the olfactory epithelium leading to delayed recovery of the epithelium. The later may explain the prolonged smell alteration linked to SARS-CoV-2 infection
Local specialists' experience and skills in animal behaviour studies: insights from wild chimpanzee field assistants
No full textInternational audienceAbstract The study of wild animal behaviour and cognition has greatly benefited from the foundational work of local specialists (LSs), particularly field assistants. In primate research, long-term studies rely on accurate identification and tracking of individuals—a skill often honed by LSs and passed on to international specialists (ISs). Despite growing recognition in publications, LSs' scientific contributions often remain undervalued. Here, we show that LSs at the Budongo Conservation Field Station (Uganda) reliably extract acoustic information (caller identity, sex and age, call components and production context) from long-distance pant hoot calls produced by wild chimpanzees. Importantly, LSs significantly outperform ISs at identifying individuals (LS accuracy = 50% (95% confidence interval (CI): 45–56%); IS accuracy = 8% (95% CI: 5–11%)), an important skill for recognizing and locating individuals in dense forests. LSs' performance was positively associated with duration of working experience. Given the limited field time of ISs (typically 1–2 years), LSs' expertise and longer commitment (mean 16.75 years) represent an essential yet underacknowledged scientific resource. Our study highlights LSs' critical role in ethological research—not only enhancing skills and data quality, but also potentially helping address both ethical (e.g. community involvement) and environmental (e.g. travel carbon footprint) challenges linked to fieldwork in remote locations
Derrière le sans-abrisme, la face cachée du mal-logement
No full texthttps://theconversation.com/derriere-le-sans-abrisme-la-face-cachee-du-mal-logement-273852Ce 22 janvier, une Nuit de la solidarité est organisée à Paris et dans d’autres villes françaises – une forme d’opération citoyenne de décompte des personnes dormant dans la rue. Celle-ci ne doit pas occulter ce que les travaux de recherche montrent depuis une quarantaine d’années : à la figure bien connue du sans-abri s’ajoutent d’autres formes moins visibles de précarité résidentielle et de mal-logement
ArchesWeather: An efficient AI weather forecasting model at 1.5° resolution
No full textInternational audienceOne of the guiding principles for designing AI-based weather forecasting systems is to embed physical constraints as inductive priors in the neural network architecture. A popular prior is locality, where the atmospheric data is processed with local neural interactions, like 3D convolutions or 3D local attention windows as in Pangu-Weather. On the other hand, some works have shown great success in weather forecasting without this locality principle, at the cost of a much higher parameter count. In this paper, we show that the 3D local processing in Pangu-Weather is computationally sub-optimal. We design ArchesWeather, a transformer model that combines 2D attention with a column-wise attention-based feature interaction module, and demonstrate that this design improves forecasting skill. ArchesWeather is trained at 1.5° resolution and 24h lead time, with a training budget of a few GPU-days and a lower inference cost than competing methods. An ensemble of four of our models shows better RMSE scores than the IFS HRES and is competitive with the 1.4° 50-members NeuralGCM ensemble for one to three days ahead forecasting. Our code and models are publicly available at https://github.com/gcouairon/ArchesWeather
Government polling in times of crises: when capacity meets incentives
No full textInternational audienceHow do governments mobilise public opinion in times of crises? While recent research examines the factors that determine the intensity of government polling at different points in the electoral cycle and the different representational logics behind this activity, empirical evidence on the more qualitative aspect of government polling is still lacking. What types of policy issues are covered in government polls? Understanding governments as actors in the production of public opinion, not just passive consumers, we examine the selection of issues in government polls. We present evidence from Germany, mobilising an original database of all poll questions directly commissioned by the German federal government during the 18th and 19th legislative periods (2013–2021). Using a conditional logit approach, we analyse how the character of the policy domain to which an issue belongs affects the likelihood that some issues are covered by government polls. Our results show that while motivations to ask questions about a particular issue are shaped by constraints (institutional, financial and political) on the government's ability to act in this area, incentives related to the salience of the issue can shift the focus of government polls to issues where they have less room for manoeuvre
Connecting mitochondrial metabolism and mitotic fidelity to control vulnerability of high grade serous ovarian cancer patients to taxane-based chemotherapy
No full textHigh-grade serous ovarian carcinoma (HGSOC), which accounts for approximately 75% of ovarian cancer cases, is associated with poor clinical outcome. Although most patients initially achieve a complete response to conventional chemotherapy, HGSOC almost invariably develops chemoresistance. There is therefore an urgent need to identify predictive biomarkers of treatment response. Here, through integrative analyses of molecular and clinical data from HGSOC patient cohorts, we identify syntabulin (SYBU), a microtubule-associated protein originally described as a regulator of mitochondrial transport along neuronal microtubules, as a critical determinant of chemosensitivity in HGSOC. Low SYBU expression in tumors correlates with higher tumor grade and increased aggressiveness, yet paradoxically with enhanced sensitivity to chemotherapy. SYBU-deficient cancer cells display impaired oxidative phosphorylation and a metabolic shift toward glycolysis characteristic of the Warburg effect, together with mitotic defects such as chromosome lagging that promote aneuploidy. Mechanistically, syntabulin forms a complex with the mitochondrial outer membrane porin VDAC1 and the inner membrane protein MIC60, a major regulator of mitochondrial cristae organization. Functionally, the syntabulin-MIC60 axis controls cristae architecture and mitotic fidelity, thereby connecting mitochondrial metabolism to cell division. These findings highlight new therapeutic vulnerabilities to overcome chemoresistance in ovarian cancer. SIGNIFICANT STATEMENT Ovarian cancer remains the deadliest gynecologic malignancy, largely due to the systematic emergence of resistance to chemotherapy. Identifying molecular mechanisms involved in response to treatment is therefore a major clinical challenge. Here, we uncover an unexpected role for the mitochondrial protein syntabulin in regulating chemotherapy sensitivity in high-grade serous ovarian cancer. We demonstrate that syntabulin coordinates cancer cell mitotic progression with mitochondrial structure and metabolism through interactions with cristae-shaping proteins. These findings reveal a previously unrecognized link between mitotic regulation and mitochondrial architecture, and identify syntabulin as a potential therapeutic target in ovarian cancer to induce vulnerability to taxane-based chemotherapy