Biology subjects

Desrosiers, P.

Publications and source records attributed to Desrosiers, P..

9 recordsLinked to original sources

bioRxiv2026

Signatures of brain-wide functional maturation in larval zebrafish

Brain development involves the coordinated organization of cellular activity into distributed functional networks, yet how this organization emerges across an entire vertebrate brain remains poorly understood. Here, we combined brain-wide structural and functional imaging to characterize brain maturation in larval zebrafish from 4 to 8 days post-fertilization. Brain growth was regionally heterogeneous and accompanied by progressive remodeling of mesoscale functional networks, with functional connectivity strengthening or weakening within distinct subnetworks. Spontaneous activity became more correlated in regions exhibiting pronounced neurogenesis, as revealed by HaloTag-based neuronal birthdating. Functional interactions became progressively less constrained by physical distance, while network organization became more modular, suggesting that long-range functional integration develops alongside increasing segregation of large-scale brain systems. At the cellular level, neuronal activity consolidated into larger, more modular functional ensembles whose combinations increasingly differentiated mesoscale modules. Visually responsive neurons redistributed along visual processing pathways, becoming more prevalent in regions further removed from retinal input, while motor-related activity declined broadly across the brain. The forebrain emerged as a major locus of maturation, exhibiting substantial anatomical expansion and a growing population of neurons displaying intrinsic activity fluctuations distinct from sensory and motor processes. Together, these findings reveal coordinated anatomical, cellular, and network-level signatures of early vertebrate brain maturation.

neuroscience↗

bioRxiv2026

Graph-based characterization of in vitro neuronal network maturation using machine learning and digital holographic microscopy

SignificanceDigital Holographic Microscopy (DHM) provides label-free quantitative phase images (QPIs) of living cells and has become a powerful tool for studying cellular morphology and dynamics. While most DHM studies have focused on cell-level analysis, the quantitative characterization of neuronal network organization and maturation from DHM images remains largely unexplored, highlighting the need for dedicated computational approaches. AimWe aimed to develop an automated framework combining deep-learning-based image analysis and graph theory to quantitatively characterize the organization, connectivity, and maturation of neuronal networks in primary rat cortical cultures imaged by DHM. ApproachTwo U-Net convolutional neural networks were trained on manually annotated DHM phase images to segment neuronal cell bodies and neurites. The resulting segmentation maps were used to infer putative morphological connections between neurons and generate graph representations of neuronal networks, referred to as graph fingerprints. A panel of 18 connectomics-inspired graph features was then computed to characterize local and global properties of network organization across four stages of culture maturation. ResultsThe mean area under the receiver operating characteristic curves was 0.98 for cell-body and 0.91 for neurite segmentation, indicating near-perfect identification. Graph-theoretical analysis revealed reproducible topological changes during network maturation in vitro, including increased density, reduced modularity, and progressive network integration. Correlation analysis showed that the 18 graph features grouped into two highly correlated families. A Random Forest classifier identified density and modularity as the most informative descriptors, achieving an accuracy of 87% in classifying maturation stages of neuronal cultures. ConclusionsOur results demonstrate that combining DHM, deep-learning-based segmentation, and graphtheoretical analysis enables quantitative characterization of neuronal network organization and maturation from label-free phase images. This framework provides a foundation for future studies of pharmacological experiments, neuronal network phenotyping, and human induced pluripotent stem cell (hiPSC)-derived neuronal cultures, where quantitative assessment of network organization remains a major challenge.

neuroscience↗

bioRxiv2026

Inferring biodiversity from indicator species using co-occurrence network structure

Biodiversity loss is accelerating worldwide, yet comprehensive assessment of species assemblages across environmental gradients remains impractical for ecological forecasting and conservation. Indicator species are therefore widely used as proxies for community composition, but existing approaches struggle to reconstruct full assemblages, particularly for rare or poorly sampled species. Here we present an integrated framework combining indicator species analysis, ecological network theory, and machine learning to infer the occurrence of non-indicator species from assemblage data. By leveraging structural properties of species co-occurrence networks, the approach captures latent community structure without explicit environmental modeling, enabling inference under sparse sampling. Across species abundance gradients, we identify three distinct regimes of predictability: high predictability for abundant species, reduced performance for species of intermediate prevalence, and unexpectedly strong predictability for rare species driven by strong co-occurrence structure with indicator species. By exploiting complementary information captured by multiple network metrics, the framework recovers species with diverse connectivity profiles and consistently outperforms richness-based or random indicator selection in both accuracy and coverage. Overall, this data-efficient approach provides a transferable pathway for biodiversity monitoring and forecasting, while offering new insights into the network organization of ecological communities.

ecology↗

bioRxiv2025

Function aligns with geometry in locally connected neuronal networks

The geometry of the brain imposes fundamental constraints on neuronal network organization and dynamics, yet how these constraints give rise to observed patterns of brain activity remains unclear. Here, we investigate how geometric eigenmodes relate to functional connectivity gradients in three-dimensional neural systems using a combination of generative network simulations and cellular-resolution calcium imaging in larval zebrafish. We show that functional connectivity gradients emerging from network activity naturally align with the geometric eigenmodes of the underlying spatial embedding when connectivity is predominantly local. By systematically increasing the prevalence of long-range connections, we reveal a robust geometry-function correspondence that progressively deteriorates as local connectivity is disrupted. We then show that spatial filtering can artificially imprint geometric patterns on functional gradients, highlighting an important methodological confound. To validate our computational results, we conduct volumetric calcium imaging experiments at cellular resolution in the optic tectum of zebrafish larvae, uncovering functional gradients that closely align with geometric eigenmodes. As predicted from simulations, the eigenmode-gradient mapping exhibits a cutoff point that quantitatively reflects the spatial extent of the regions connectivity kernel, inferred from single-neuron morphologies. This geometry-functional alignment disappears at brain-wide scale, where long-range connections are more prevalent. Our findings demonstrate how short-range anatomical connectivity anchors functional connectivity gradients to the brains geometry.

neuroscience↗

bioRxiv2024

NeuroTorch: A Python library for neuroscience-oriented machine learning

Machine learning (ML) has become a powerful tool for data analysis, leading to significant advances in neuroscience research. While ML algorithms are proficient in general-purpose tasks, their highly technical nature often hinders their compatibility with the observed biological principles and constraints in the brain, thereby limiting their suitability for neuroscience applications. In this work, we introduce NeuroTorch, a comprehensive ML pipeline specifically designed to assist neuroscientists in leveraging ML techniques using biologically inspired neural network models. NeuroTorch enables the training of recurrent neural networks equipped with either spiking or firing-rate dynamics, incorporating additional biological constraints such as Dales law and synaptic excitatory-inhibitory balance. The pipeline offers various learning methods, including backpropagation through time and eligibility trace forward propagation, aiming to allow neuroscientists to effectively employ ML approaches. To evaluate the performance of NeuroTorch, we conducted experiments on well-established public datasets for classification tasks, namely MNIST, Fashion-MNIST, and Heidelberg. Notably, NeuroTorch achieved accuracies that replicated the results obtained using the Norse and SpyTorch packages. Additionally, we tested NeuroTorch on real neuronal activity data obtained through volumetric calcium imaging in larval zebrafish. On training sets representing 9.3 minutes of activity under darkflash stimuli from 522 neurons, the mean proportion of variance explained for the spiking and firing-rate neural network models, subject to Dales law, exceeded 0.97 and 0.96, respectively. Our analysis of networks trained on these datasets indicates that both Dales law and spiking dynamics have a beneficial impact on the resilience of network models when subjected to connection ablations. NeuroTorch provides an accessible and well-performing tool for neuroscientists, granting them access to state-of-the-art ML models used in the field without requiring in-depth expertise in computer science.

neuroscience↗

bioRxiv2024

Structural and genetic determinants of zebrafish functional brain networks

Network science has significantly advanced our understanding of brain networks across species, revealing universal connectivity principles. While human studies based on magnetic resonance imaging (MRI) have established several network principles at macroscopic scales, recent breakthroughs, including high-amplitude regional co-activation patterns and spatially contiguous functional gradients, remain unexplored at cellular resolution in animal models. Here, we employ whole-brain functional imaging at cellular resolution in larval zebrafish, combined with anatomical and spatial genetic expression profile databases, to investigate the structural and genetic basis of functional brain networks. We show that mesoscopic functional connectivity (FC) is a robust measure of brain activity that captures the individuality of larvae. Using a public dataset of thousands of single-neuron reconstructions, we reveal a strong coupling between FC and structural connectivity (SC). Numerous properties of the connectome that account for indirect pathways and diffusion mechanisms individually and collectively predict interregional correlations. The hierarchical modular structure of SC and FC significantly overlaps in space, and modules identified within the connectome constrain the shape of both spontaneous and stimulus-driven activity patterns. Using visual stimuli and tail monitoring, we identify a functional network gradient that maps onto the sensorimotor function of brain regions. Finally, we identify a set of genes whose co-expression in brain regions significantly predicts regional FC. Our findings reproduce several key features of mammalian brain networks in zebrafish, demonstrating the potential for studying large-scale network phenomena in smaller, optically accessible vertebrate brains.

neuroscience↗

bioRxiv2023

A probabilistic methodology to reconstruct biodiversity from implicit interactions with indicator species along latitudinal co-occurrence networks

The relationship between species presence, biodiversity reconstruction, and latitudinal gradients is a complex and multifaceted topic that has been the subject of extensive research in ecology. Recent studies have provided valuable insights into the patterns and drivers of these phenomena. Also, with the ongoing decline in biodiversity, there is a need for efficient field monitoring techniques. Indicator species (IS) emerged as a promising tool to monitor diversity because their presence indicates a maximum number of conditionally co-occurring species. We aim to assess the effectiveness of IS for biodiversity reconstruction implicitly based on their co-occurrence with other species through a network-based methodology. The IS are identified based on various network metrics and the likelihood of species occurrences is computed based on (1) their conditional occurrence probability with IS and (2) the occurrence probability of IS. We test the approach with field observations of birds in the Cote-Nord region of Quebec. From our methodology, the climate latitudinal gradient plays a significant role on the alternation in composition of IS with an almost complete turnover between northern and southern networks. The latitudinal gradient impacts also the nature of the inter-specific interactions with more avoidance relationship toward the Tropics and more cooperation liaisons toward the north. Regarding the effectiveness in the reconstruction of assemblages occurrence, we observe a strong negative correlation (r [≤] -0.95) between the percentage of sites occupied and the dissimilarity between the original and the estimated occurrences. More precisely, species must be present in more than 29% and 33% of northern and southern sites to recover well from its co-occurrence with IS. Therefore, it is more challenging to reconstruct biodiversity in communities closet to Tropics due to higher complex interactions and interspecific competition in these areas, which make it more difficult to infer community composition. In conclusion, our method demonstrates that it is possible to predict local species assemblages based on their implicit interactions with local IS. Nevertheless, the relatively low success of less present species illustrates the need for further theoretical development to reconstruct biodiversity, mainly to recover the occurrence of rare species.

ecology↗

bioRxiv2022

Long-term effect of forest harvesting on boreal species assemblages under climate change

Logging is the main human disturbance impacting biodiversity in forest ecosystems. However, the impact of forest harvesting on biodiversity is modulated by abiotic conditions through complex relationships that remain poorly documented. Therefore, the interplay between forest management and climate change can no longer be ignored. Our aim was to study the expected long-term variations in the assemblage of bird and beetle communities following modifications in forest management under different climate change scenarios. We developed species distribution models to predict the occurrence of 87 species of birds and beetles in eastern Canadian boreal forests over the next century. We simulated three climate scenarios (baseline, RCP4.5 and RCP8.5) under which we varied the level of harvesting. We also analyzed the regional assemblage dissimilarity by decomposing it into balanced variations in species occupancy and occupancy gradient. We predict that forest harvesting will alter the diversity by increasing assemblage dissimilarity under all the studied climate scenarios, mainly due to species turnover. Species turnover intensity was greater for ground-dwelling beetles, probably because they have lower dispersal capacity than flying beetles or birds. A good dispersal capacity allows species to travel more easily between ecosystems across the landscape when they search for suitable habitats after a disturbance. Regionally, an overall increase in the probability of occupancy is projected for bird species, whereas a decrease is predicted for beetles, a variation that could reflect differences in ecological traits between taxa. Our results further predict a decrease in the number of species that increase their occupancy after harvest under the most severe climatic scenario for both taxa. We anticipate that under severe climate change, increasing forest disturbance will be detrimental to beetles associated with old forests but also with young forests after disturbances.

ecology↗

bioRxiv2022

Effects of global change on animal biodiversity in boreal forest landscape: an assemblage dissimilarity analysis

Despite an increasing number of studies highlighting the impacts of climate change on boreal species, the main factors that will drive changes in species assemblages remain ambiguous. We quantify two climate-induced pathways based on direct and indirect effects on species occupancy and assemblage dis-similarity under different harvest management scenarios. The direct climate effects illustrate the impact of climate variables while the indirect effects are reflected through the changes in land cover composition. To understand the main causes in assemblage dissimilarity, we analyze the regional and the latitudinal species assemblage dissimilarity by decomposing it into balanced variation in species occupancy and occurrence and occupancy and occurrence gradient. We develop empirical models to predict the distribution of more than 100 bird and beetle species in the Cote-Nord region of Quebec over the next century. Our results show that the two pathways are complementary and alter biodiversity, mainly caused by balanced variation in species occupancy and occurrence. At the regional scale, both effects have an impact on decreasing the number of winning species. Yet, responses are much larger in magnitude under mixed climate effects (a mixture of direct and indirect effects). Regional assemblage dissimilarity reached 0.77 and 0.69 under mixed effects versus 0.09 and 0.10 under indirect effects for beetles and birds, respectively, between RCP8.5 and baseline climate scenarios when considering harvesting. Therefore, inclusion of climatic variables considers aspects other than just those related to forest landscapes, such as life cycles of animal species. Latitudinally, assemblage dissimilarity increased following the climate conditions pattern. Our analysis contributes to the understanding of how climate change alters biodiversity by reshaping community composition and highlights the importance of climate variables in biodiversity prediction.

ecology↗