Archives ouvertes de Paris-Saclay
Not a member yet
272730 research outputs found
Sort by
Ecological drivers of intercropping performance for enhanced global crop production
No full textSource Agritrop Cirad (https://agritrop.cirad.fr/616391/) * Autres projets (id;sigle;titre): 101082057;TC4BE;(EU) Transformative Change in Telecoupled Agrofood Systems for Biodiversity and Equity//International audienceIntercropping, a cornerstone of ecological intensification for sustainable agriculture and biodiversity conservation, has not fulfilled its potential to support global food-security promises under growing land and climate constraints. A major barrier lies in the context-dependent and often unpredictable nature of yield benefits, which emerge from complex interspecific interactions that remain poorly characterized across agroecosystems. Through a novel global meta-analysis (4,195 partial Land Equivalent Ratio observations from 334 studies across 60 countries) coupled with machine learning to disentangle the roles of plant functional traits and species-specific interactions, we demonstrate a substantial untapped potential to increase the production of major cereals—maize ( + 51%), barley ( + 6%), and wheat ( + 1%)—solely through optimized deployment of intercropping on existing agricultural land. Crucially, despite reduced planting density in intercropping, the mean partial Land Equivalent Ratio (pLER) of 0.79 (95% CI: 0.76-0.82) indicates that yield reductions for component species are proportionally smaller than the decrease in density, revealing consistent beneficial interactions. Our quantitative framework identifies relative planting density (RD), temporal niche differentiation (TND), and relative height difference as key levers for optimizing intercropping performance. We further unveil a predictable trade-off governed by asymmetric competition: targeted manipulation of RD and TND can selectively benefit either taller or shorter species, providing mechanistic insights into how interspecific dynamics shape intercropping success. This finding offers a scalable and ecologically grounded pathway to increase global crop production without cropland expansion, advancing sustainable agricultural intensification
Aligning the Unseen in Attributed Graphs: Interplay between Graph Geometry and Node Attributes Manifold
No full textThe standard approach to representation learning on attributed graphs---i.e., simultaneously reconstructing node attributes and graph structure---is geometrically flawed, as it merges two potentially incompatible metric spaces. This forces a destructive alignment that erodes information about the graph’s underlying generative process. To recover this lost signal, we introduce a custom variational autoencoder that separates manifold learning from structural alignment. By quantifying the metric distortion needed to map the attribute manifold onto the graph’s Heat Kernel, we transform geometric conflict into an interpretable structural descriptor. Experiments show our method uncovers connectivity patterns and anomalies undetectable by conventional approaches, proving both their theoretical inadequacy and practical limitations
Le projet de recherche S-PASS : ressources et usages du sous-sol de la métropole du Grand Paris
No full textInternational audienceFor the city, the subsoil represents fundamental challenges for its development and evolution. It is the foundation for surface developments, a place that houses public transportation and technical networks. It is also a source of geothermal energy that is still insufficiently exploited in urban areas. In the context of climate change, the effects of which will be particularly marked over the coming decades in major metropolises, the subsoil can also become a new space to be developed as an alternative to urban sprawl. The S-PASS scientific research project, part of the “Subsurface: a common good” research program funded under the France 2030 plan, is built around these issues, on the geographical perimeter of the Greater Paris metropolis. The project covers the first 200 m of the subsoil, and aims to (1) gain a better understanding of the geological formations, their associated variability, and their geomechanical properties, and to test new geophysical methods in urban areas; (2) build a 3D digital model coupling the 3D geological model of the Cenozoic geological formations, with existing underground public transport infrastructures. The merging of these two models will enable the creation of a digital twin prototype of the Parisian urban underground; (3) to place these underground spaces in the public imagination and in its perception of future developments, to analyse the environmental footprint of underground use in the urban development model in comparison with surface developments ; and finally (4) to consider innovative methodologies for increasing circular economy applications, i.e. reclaiming excavated soil from underground works and using low enthalpy geothermal solutions. This research project, with a budget of 3 million euros and a duration of 7 years, started in 2023. It brings together eleven academic and institutional partners. This article describes the stakes, content and prospects of this project.Le sous-sol représente pour la ville des enjeux fondamentaux pour son développement et son évolution. Il est à la fois le support des fondations des aménagements de surface, un lieu qui abrite les transports publics et les réseaux. Il contient également une source d'énergie géothermique encore insuffisamment exploitée en zone urbaine. Dans le contexte de changement climatique, le sous-sol peut aussi devenir un nouvel espace à développer, en alternative de l'étalement urbain. Le projet de recherche scientifique S-PASS, inscrit dans le programme de recherche PEPR « Sous-sol -Bien commun », financé dans le cadre du plan France 2030, est construit autour de ces enjeux, sur le périmètre géographique de la métropole du Grand Paris. Il s'intéresse aux premiers 200 m du sous-sol et se décline en quatre objectifs. Le premier consiste à mieux connaitre les formations géologiques, leur variabilité spatiale et leurs propriétés géomécaniques. Le second objectif vise à construire une maquette numérique 3D couplant d'une part le modèle géologique 3D des formations cénozoïques et d'autre part, les infrastructures souterraines existantes de métro, afin de créer un prototype de jumeau numérique du sous-sol parisien. Le troisième objectif est dédié à la durabilité et à la perception sociétale des villes souterraines. Il vise d'une part à replacer ces espaces souterrains dans l'imaginaire du grand public et dans sa perception des développements futurs et d'autre part à construire une méthodologie permettant une qualification environnementale objective de l'utilisation du sous-sol en comparaison d'aménagements de surface. Enfin le dernier volet, focalisé sur les ressources souterraines, vise à envisager des méthodologies innovantes pour la valorisation des terres excavées, et à étudier la géothermie de proche surface. Ce projet de recherche, doté d'un budget de 3 millions d'euros pour une durée de 7 ans, a démarré en 2023. Il réunit onze partenaires académiques et institutionnels. Cet article détaille les enjeux, le programme et les perspectives du projet
Euclid: Early Release Observations -- The extended stellar component of the IC10 dwarf galaxy
No full textInternational audienceWe present a detailed analysis of the old, extended stellar component of the Local Group dwarf galaxy IC 10 using deep resolved-star photometry in the VIS and NISP bands of the Euclid Early Release Observations. Leveraging Euclid's unique combination of wide field of view and high spatial resolution, we trace red giant branch (RGB) stars out to 8 kpc from the galaxy centre, reaching azimuthally-averaged surface brightness levels as faint as 29 mag arcsec. Our analysis reveals that IC 10's stellar distribution is significantly more extended than previously thought. After correcting for foreground extinction and subtracting contamination from Milky Way stars and background galaxies, we derive a radial stellar density profile from RGB star counts. The profile shows a marked flattening beyond 5 kpc, and is best fit by a two-component (Sersic + exponential) model, yielding a total stellar mass in old (age 1 Gyr) stars of . The origin of the outer stellar component is unclear. It may be accreted, even possibly associated with the counter-rotating HI gas in the outer regions of IC 10, or it may represent an ancient in-situ stellar halo. We tentatively detect two symmetric stellar overdensities at the edge of our imagery. These roughly align with the direction of IC 10's orbit around M31, suggesting that they may be signatures of tidal stripping. As part of our analysis, we derive a new distance to IC 10 based on the RGB tip, finding kpc and the distance modulus is
Euclid preparation. Calibrated intrinsic galaxy alignments in the Euclid Flagship simulation
No full textInternational audienceIntrinsic alignments of galaxies are potentially a major contaminant of cosmological analyses of weak gravitational lensing. We construct a semi-analytic model of galaxy ellipticities and alignments in the \Euclid Flagship simulation to predict this contamination in Euclid's weak lensing observations. Galaxy shapes and orientations are determined by the corresponding properties of the host haloes in the underlying -body simulation, as well as the relative positions of galaxies within their halo. Alignment strengths are moderated via stochastic misalignments, separately for central and satellite galaxies and conditional on the galaxy's redshift, luminosity, and rest-frame colour. The resulting model is calibrated against galaxy ellipticity statistics from the COSMOS Survey, selected alignment measurements based on Sloan Digital Sky Survey samples, and galaxy orientations extracted from the Horizon-AGN hydrodynamic simulation at redshift . The best-fit model has a total of 12 alignment parameters and generally reproduces the calibration data sets well within the statistical uncertainties of the observations and the \flagship simulation, with notable exceptions for the most luminous sub-samples on small physical scales. The statistical power of the calibration data and the volume of the single \flagship realisation are still too small to provide informative prior ranges for intrinsic alignment amplitudes in relevant galaxy samples. As a first application, we predict that \Euclid end-of-mission tomographic weak gravitational lensing two-point statistics are modified by up to order due to intrinsic alignments
Factorisation matricielle non négative parcimonieuse pour le traitement des données de spectrométrie de masse en métabolomique
No full textMass spectrometry (MS) based metabolomics is a leading strategy for biomarker discovery. In particular, tandem mass spectrometry facilitates the structural characterization of metabolites through the acquisition of fragmentation spectra (MS2). Recent approaches to Data Independent Acquisition (DIA) such as SWATH (Sequential Window Acquisition of all THeoretical fragment ions) enable high-throughput fragmentation of all precursor ions. However, the SWATH DIA spectra obtained are hybrid, as they contain mixed fragments from all co-isolated compounds. This is why the development of strategies capable of separating these spectra in a completely untargeted manner is a major challenge for metabolite identification. To process SWATH DIA data rigorously and efficiently, we have developed a new approach based on sparse non-negative matrix factorization (NMF) named DIANMF. This is the first blind source separation method that simultaneously processes MS1 and MS2 data in a unified framework, without relying on predefined peak models or spectral libraries. We have implemented an NMF algorithm derived from non-negative generalized morphological analysis with soft thresholding (nGMCAˢ), which exploits non-negativity and parsimony (in a predefined known transformation domain) as well as morphological diversity to improve separation in heavily mixed linear data. We have shown through simulation that a high rank can be chosen a priori so that the spectra of all compounds present in the time window can be extracted simultaneously at each iteration, and we have developed criteria for selecting only the components corresponding to signal and assigning each to the precursor ions of the same molecule. The entire DIANMF workflow is implemented as the publicly available software library DIANMF (https://github.com/odisce/DIANMF). Application of the software to several real-world datasets representative of the main detection instruments has shown that DIANMF outperforms existing tools in terms of compound identification rates and spectral quality, particularly for low-abundance and co-eluting metabolites. Furthermore, our approach using simultaneous NMF on MS1 and MS2 data makes it possible for the first time to group all ion species associated with the same compound (adducts, isotopes, and neutral losses) into a chemically coherent spectrum. Taken together, these results demonstrate the value of DIANMF for the processing of SWATH DIA data and the identification of biomarkers in metabolomics.La métabolomique par spectrométrie de masse (MS) est une approche majeure pour la découverte de biomarqueurs. En particulier, l'acquisition de spectres de fragmentation (MS2) permet de caractériser la structure des métabolites. Les approches récentes d'acquisition indépendante des données (DIA) de type SWATH (Sequential Window Acquisition of all THeoretical fragment ions) offrent la possibilité de fragmenter tous les ions détectés. Toutefois, les spectres SWATH DIA obtenus sont des mélanges de tous les fragments provenant des composés co-isolés. Pour traiter les données SWATH DIA de manière rigoureuse et efficace, nous avons développé une nouvelle approche reposant sur la factorisation matricielle non négative (NMF) parcimonieuse, baptisée DIANMF. Il s'agit de la première méthode de séparation de sources en aveugle qui traite simultanément les données MS1 et MS2 dans un cadre unifié, sans s'appuyer sur des modèles de pics prédéfinis ou des bibliothèques spectrales. Nous avons implémenté un algorithme NMF dérivé de l'analyse morphologique généralisée non négative avec seuillage doux (nGMCAˢ), qui exploite la non-négativité et la parcimonie (dans un domaine de transformation connu prédéfini) ainsi que la diversité morphologique pour améliorer la séparation dans les données linéaires fortement mélangées. Nous avons montré par simulation qu'un rang élevé peut être choisi a priori de manière à pouvoir extraire simultanément les spectres de tous les composés présents dans la fenêtre temporelle à chaque itération, et nous avons développé des critères pour sélectionner uniquement les composantes correspondant à du signal et d'attribuer chacune aux ions précurseurs d'une même molécule. L'ensemble de la méthodologie est implémentée sous la forme de la librairie logicielle publiquement disponible DIANMF (https://github.com/odisce/DIANMF). L'application du logiciel à plusieurs jeux de données réels représentatifs des principaux instruments de détection a montré que DIANMF obtient de meilleures performances que les outils existants en termes de taux d'identification des composés et de qualité spectrale, notamment pour les métabolites à faible abondance et co-élués. En particulier, notre approche permet pour la première fois de regrouper toutes les espèces ioniques associées à un même composé (adduits, isotopes et pertes neutres) dans un spectre chimiquement cohérent. L'ensemble de ces résultats montrent l'intérêt de DIANMF pour le traitement des données SWATH DIA et l'identification des composés en métabolomique
Identification of receptor-binding domains of Bacteroidales antibacterial pore-forming toxins
No full textInternational audienceBacteroidales are abundant Gram-negative bacteria present in the gut microbiota of most animals, including humans, where they carry out vital functions for host health. To thrive in this competitive environment, Bacteroidales use sophisticated weapons to outmatch competitors. Among these, BSAPs (Bacteroidales Secreted Antimicrobial Proteins) represent a novel class of bactericidal pore-forming toxins that are highly specific to their receptor, typically targeting only a single membrane protein or lipopolysaccharide. The molecular determinants conferring this high selectivity remain unknown. In this study, we therefore investigated the model protein BSAP-1 and determined which of its domains is involved in providing receptor specificity. We demonstrate that receptor recognition is entirely driven by the C-terminal domain (CTD) of BSAP-1 using a combination of in vivo competition assays, in vitro protein binding studies and mutational analysis. Specifically, we show that deletion of the CTD abrogates BSAP-1 bactericidal activity by preventing receptor binding, while grafting the CTD to unrelated carrier proteins enables CTD-driven interaction with the BSAP-1 receptor. Combining structural investigation of a BSAP-1-receptor complex with mutational analysis, we unveil that this interaction is driven by electrostatic interactions. Building upon this discovery, we show that BSAPs can be categorized according to the structure of their CTD, suggesting a strong CTD structure/receptor type correlation. In summary, our research demonstrates that BSAP receptor recognition is driven by their CTD and paves the way for future applications
A multi-omic resource for exploring microbial eukaryotes in the meromictic freshwater Lake Pavin
No full textInternational audienceAlthough recent advances in high-throughput sequencing have greatly expanded our understanding of microbial diversity and function in aquatic ecosystems, progress in studying freshwater microbial eukaryotes has been more limited, mainly due to their large genomes, immense diversity, and largely uncharacterised physiologies. In this work, we present a comprehensive multi-omic dataset, eukaryote-centred, including targeted-metagenomic (18S rDNA V4 and V9), metagenomic, metatranscriptomic and single amplified genomes (SAGs). Both the oxic and anoxic layers of Lake Pavin (France), a permanently stratified freshwater lake, were sampled at four distinct times throughout 2018, by day and night, targeting microbial eukaryotes of two size classes (0.65–10 µm and 10–50 µm). This dataset comprises 106 eukaryotic metagenome-assembled genomes (MAGs), over 9 million unigenes and 11 SAGs, encompassing several under-represented taxa in public databases ( e.g . Perkinsea, Chytridiomycota, Cryptista). Altogether, this dataset represents a resource for exploring the functional diversity and spatio-temporal dynamics of microbial eukaryotes
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
Adaptive federated control: An event-driven MARL framework for fair and efficient traffic management
No full textInternational audienceNext-generation networks require distributed traffic management to handle dynamic loads, but frequent interagent coordination consumes scarce bandwidth. We propose an adaptive federated multi-agent reinforcement learning (Fed-MARL) framework that triggers model synchronization only when congestion is near. Our congestion index combines queue occupancy, latency, and utilization to detect network stress. When the threshold is exceeded, federated learning aggregates distributed agent models using FedAdam. For fairness, we introduce empathy-weighted reward shaping, where agents balance individual rewards with peer performance, aligning selfish routing with system-wide welfare. DDPG agents deployed at edge switches make routing and load balancing decisions. Evaluated on Fat-Tree K=4 topology, Fed-MARL achieves 93% latency reduction vs. RL-MR (7.77 vs 105 ms), 57% vs. DRAMA (18.15 ms), 192.6 Mbps throughput, and perfect delivery with minimal communication overhead