HAL Evry
Not a member yet
19237 research outputs found
Sort by
Le rôle des festivals dans les transferts culturels transatlantiques
No full textInternational audienc
MCQR: Enhancing the processing and analysis of quantitative proteomics data by incorporating chromatography and mass spectrometry information
No full textInternational audienceIn the field of proteomics, generating biologically relevant results from mass spectrometry signals remains a challenging task. This is partly due to the fact that the computational strategies for converting MS signals into biologically interpretable data depends heavily on the MS acquisition method. Additionally, the processing and the analysis of these data vary depending on whether the proteomic experiment was performed with or without labeling, and with or without fractionation. Several R packages have been developed for processing and analyzing MS data, but they only incorporate identification and quantification data; none of them takes into account other invaluable information collected during MS runs. To address this limitation, we introduce MCQR, an alternative R package for the in-depth exploration, processing, and analysis of quantitative proteomics data
Contributions à l’apprentissage automatique pour la calibration des capteurs de pollution de l’air à faible coût
No full textThis thesis proposes frugal deep learning models that utilize direct and indirect calibration approaches to improve sensor accuracy. We begin with a comprehensive review of the state-of-the-art, outlining existing techniques and their limitations in handling domain shift, noise, and environmental variability. We then develop a direct calibration strategy based on linear and non-linear models. Our findings reveal that while linear models may suffice under controlled conditions, non-linear are essential in complex outdoor environments to account for pollutant interferences and environmental fluctuations. Furthermore, we introduce an indirect calibration approach that solves domain adaptation. We align latent representations between source and target domains by combining variational with adversarial unsupervised training. This enhances calibration models' robustness and generalizability. We follow that with comparative studies that demonstrate through real-world situations that although direct calibration achieves satisfactory results in simpler scenarios, domain adaptation is needed to address significant domain shifts.Dans cette thèse, nous proposons des modèles d'apprentissage profond sobres qui exploitent à la fois des approches directes et indirectes pour calibrer les capteurs de polluants de l'air à faibles coût. Nous commençons par présenter une revue exhaustive de l'état de l'art, en exposant les techniques existantes et leurs limites face au bruit et à la variabilité environnementale. Nous développons ensuite une stratégie de calibration directe reposant sur des modèles linéaires et non linéaires. Nos résultats montrent qu'en environnement contrôlé, un modèle linéaire peut être suffisant pour établir la relation entre les capteurs et les données de référence. En revanche, dans des conditions extérieures complexes, où les interférences entre polluants et les fluctuations environnementales sont plus marquées, l'utilisation de modèles non-linéaires s'avère indispensable. Par ailleurs, nous proposons une approche de calibration indirecte basée sur l'adaptation de domaines. En intégrant des auto-encodeurs variationnels et des techniques d'apprentissage adversariales, nous alignons les représentations latentes entre le domaine source et le domaine cible, renforçant ainsi la robustesse et la généralisabilité des modèles. Les études comparatives indiquent que, bien que la calibration directe offre des résultats satisfaisants dans des scénarios simples, l'adaptation de domaines est essentielle pour traiter efficacement des décalages importants dans des environnements réels
A new deep-branching environmental lineage of algae
No full textAbstract Marine algae support the entire ocean ecosystem and beyond. Algae in culture poorly represents their large environmental diversity, and we still have a limited understanding of their convoluted evolution by endosymbiosis. Here, we performed a phylogeny-guided plastid genome-resolved metagenomic survey of Tara Oceans expeditions. We present a curated resource of 660 new non-redundant plastid genomes of environmental pelagic algae. This catalogue vastly expands the plastid genome diversity within major algal groups, often corresponding to algae without closely related reference genomes. Notably, we recovered four plastid genomes, including one near complete, forming a deep-branching plastid lineage of nano-size algae that we informally named leptophytes. This group is globally distributed and generally rare, although it can reach relatively high abundance at least in the Arctic. A mitochondrial contig including 62 genes showing strong read coverage correlation with leptophytes was also recovered from these Arctic samples and assigned to this group. Leptophytes encompass the enigmatic marine plastid group DPL2, one of the very few known plastid groups not clearly belonging to any major algal groups and for which only 16S rDNA amplicon data is available. Extensive gene content comparison and organellar phylogenomics support the view that leptophytes are sister to haptophytes, and raise the intriguing possibility that cryptophytes acquired their plastids from haptophytes. Collectively, our study demonstrates that metagenomics can reveal currently hidden diversity of organellar genomes, and shows the importance of including this diversity to improve models for plastid evolution
Decoding transcriptional identity during neuron-astroglia cell fate driven by RAR-specific agonists
No full textInternational audienceHow cells respond to different signals leading to defined lineages is an open question to understand physiological differentiation leading to the formation of organs and tissues. Among the various morphogens, retinoic acid signaling, via the RXR/RAR nuclear receptors activation, is a key morphogen of nervous system development and brain homeostasis. Here, we analyze gene expression in ∼80 000 cells covering 16 days of monolayer mouse stem cell differentiation driven by the pan-RAR agonist all-trans retinoic acid, the RARα agonist BMS753 or the activation of both RARβ and RARγ receptors (BMS641 + BMS961). Furthermore, we have elucidated the role of these retinoids for driving nervous tissue formation within 90 days of brain organoid cultures, by analyzing >8000 distinct spatial regions over 28 brain organoids. Despite a delayed progression in BMS641 + BMS961, RAR-specific agonists led to a variety of neuronal subtypes, astrocytes, and oligodendrocyte precursors. Spatially resolved transcriptomics performed in organoids revealed spatially distinct RAR isotype expression leading to specialization signatures associated with matured tissues, including a variety of neuronal subtypes, retina-like tissue structure signatures and even the presence of microglia
Transposable elements are vectors of recurrent transgenerational epigenetic inheritance
No full textInternational audienceDNA methylation loss at transposable elements (TEs) can affect neighboring genes and be epigenetically inherited in plants, yet the determinants and significance of this additional system of inheritance are unknown. Here, we demonstrate in Arabidopsis thaliana that transgenerational stability of experimentally-induced hypomethylation at TE loci is constrained by small RNAs derived from related copies. Using data from >700 strains collected worldwide, we uncover similar and recurrent hypomethylation at hundreds of these TE loci, often near genes. Most natural epivariants we tested can be inherited without DNA sequence changes and are therefore bona fide epialleles, although genetic factors modulate their recurrence or persistence. Epiallelic variants often cause gene expression changes and may be targets of selection, thus revealing their contribution to heritable phenotypic variation in nature
Identification of a sex-determining region potentially involved in resolving genetic conflicts over sex ratio
No full textInternational audienceSex-determining genes remain largely uncharacterized outside classical models in vertebrates and insects, leaving a gap in our understanding of their evolutionary emergence. The terrestrial isopod Armadillidium vulgare provides an excellent model for investigating this question, as it presents multiple genetic sex determinants. Some lineages possess a masculinizing dominant allele at a locus called the ‘ M gene’, which is analogous to an XY system. This allele is hypothesized to have been selected due to the deficit of males caused by a non-Mendelian feminizing factor previously described. The existence of the M gene was inferred from crosses carried out in the 1990s, but its molecular nature remains unresolved. Here, we conducted a genome-wide single-nucleotide polymorphisms analysis combining pooled sequencing of male and female progenies with sequencing of individual parents across two families. Bayesian estimation of haplotype frequencies in progenies pinpointed a candidate genomic region of approximately two megabases. Notably, this region contains the gene encoding the androgenic gland hormone, a protein involved in male sexual differentiation. Our findings lay the groundwork for functional investigations of the M gene, offering novel insights into the dynamics of sex determination in terrestrial isopods and into the turnover of sex chromosomes in response to sex-ratio distortion
SI-referenced formic acid (HCOOH) spectroscopy at the sub ppt level
No full textWe report on a new determination of the ν6, J′ = 21, K′a = 2, K′c = 20 ← J′′ =21, K′′a = 3, K′′c = 19 HCOOH rovibrational line centroid at 9.17 μm with 7.1 Hzstatistical uncertainty (1.3 10−13) , a 100-fold improvement with respect to pre-vious measurements. It involves a Fabry-Perot saturated-absorption spectrometerreferenced to the Système International through a frequency comb. We elaborate adetailed line shape model that allows retrieving the transit time width and accuratedetermination of the pressure and power shifts and broadenings. We also give theair pressure shift as well as the nitrogen and air broadening coefficients
The origins of a mathematical heritage: The library of Gaston Darboux’s cabinet of Higher Geometry (1900–1917)
No full textInternational audienceBetween 1900 and 1917, Gaston Darboux (1842-1917) built up a library associated with his Chair in Higher Geometry at the Paris Faculty of Science, which forms the initial nucleus of today’s library at the Institut Henri Poincaré, inaugurated in 1928. Created in the wake of a series of reforms instituting a new university system in France, this library serves not only as a current working tool for his teaching and research activities, but also as a testimony to the mathematics of his time
Spatio-Temporal Hyperbolic Aggregation Neural Network for Human Action Recognition
No full textInternational audienceHuman action recognition (HAR) is a critical task in the field of robotics. Traditionally, HAR methods rely on either perceptual features from RGB images or skeletal features. While RGB-based features are typically represented in 2D Euclidean space, few approaches differentiate between methods developed for RGB data and those for skeletal features, often treating both as Euclidean representations. This conventional approach, which typically leverages standard deep learning techniques, limits the descriptive power of skeletal data, which naturally exhibits a tree-like structure. In this paper, we introduce a novel framework that, for the first time, utilizes skeletal data while preserving its inherent structure to fully capture its descriptive potential. Our proposed deep neural network embeds skeletal joints into hyperbolic space, followed by a spatio-temporal processing framework that incorporates established transformations to optimize performance while maintaining the advantages of hyperbolic analysis. Extensive experiments on publicly available datasets, including UAV-Human, UAV-Gesture, and DHG 14/28, demonstrate that our approach achieves state-of-the-art results, underscoring its ability to enhance robotic systems’ performance in dynamic environments