Biology subjects

Dray, S.

Publications and source records attributed to Dray, S..

10 recordsLinked to original sources

bioRxiv2026

Six decades of change across all North American bird interaction networks

Species interactions are essential for maintaining biodiversity in the face of global changes. Yet lack of interaction data has hindered progress in understanding spatiotemporal changes in ecological network structure and function. To address this shortfall, we perform network analysis on a continental-scale metanetwork of bird interactions (AvianMetaNetwork), and quantify shifts in trophic, commensalism, parasitism, competition, mobbing, and facilitation networks using 13762 interactions among 687 North American birds. Since 1970, changes in species abundance and human activity drove network shifts, especially in eastern North America. With this metanetwork, we uncovered more than a half century of previously undocumented shifts in the structure of each avian network at a continental scale, with large implications for the functioning of bird communities and the ecosystems that rely on them.

ecology↗

bioRxiv2025

Paying Attention to Other Animal Detections Improves Camera Trap Classification Models

O_LIIn ecological studies, automated species classification models are increasingly used to process large volumes of camera trap images. Most current classification models rely on a two-step pipeline: a detector first locates and crops animals, followed by a classifier that predicts species independently for each crop. While effective, these models still struggle under challenging conditions and ignore temporal context or information from nearby animals available in sequences of camera trap images, whereas human annotators often use it in difficult cases. C_LIO_LIWe propose to leverage self-attention, a core mechanism of Large Language Models and Vision Transformers, to enable the model to learn relationships between crops occurring in similar contexts. Our self-attention module operates directly on the set of crop embeddings, producing new representations enriched with information from other crops, improving species classification. The module fits into the two-step pipeline without requiring structural change and adds only minimal computational overhead. To address the lack of annotated multi-species sequences, we develop a training strategy that synthetically generates multi-species sequences from monospecific ones. C_LIO_LICompared to an independent classifier baseline, our multi-crop model achieves higher accuracy on mono-specific sequences, both real and synthetic. In multispecies settings, evaluated using synthetic test sets, we also observe a substantial improvement in accuracy. Using real but weakly annotated multi-species sequences, we reformulate the task as multi-label set classification, and conduct a visual analysis, to highlight the benefits of our approach. C_LIO_LIBy leveraging the information brought by all detections of animals in the image and others of the same sequence, our approach reduces species misclassifications and enables more accurate estimates for downstream ecological analyses focusing on, for instance, species richness, occupancy, and species interaction. C_LI

ecology↗

bioRxiv2024

Trait matching without traits: using correspondence analysis to analyze the latent structure of interaction networks

Species interactions in ecological communities are often represented as networks, the structure of which is thought to be linked to species interaction niches (or Eltonian niches). Interaction niches are intimately related to the notion of trait matching, which posits that a species interacts preferentially with partners whose traits are complementary to their own. Multivariate methods are commonly used to quantify species environmental niches (or Grinnellian niches). More recently, some of these methods have also been used to study the interaction niche, but they consider only the niche optimum and require trait data. In this article, we use the correspondence analysis (CA) framework to study interaction networks and investigate trait matching without requiring trait data, using the notion of latent traits. We use reciprocal scaling, a method related to CA, to estimate niche optima and breadths, defined respectively as the mean and standard deviation of the latent traits of species interacting partners. We present the method, test its performance using a simulation model we designed, and analyze a real frugivory network between birds and plants. The simulation study shows that the method is able to recover niche breadths and optima for data generated with parameters typical of ecological networks. The birds-plants network analysis shows strong relationships between species latent traits and niche breadths: a posteriori correlation with measured traits suggests that birds and plants of intermediate size tend to have the broadest niches. Additionally, birds preferentially foraging in the understory have broader niches than birds preferentially foraging in the canopy. CA and reciprocal scaling are described as fruitful exploratory methods to characterize species interaction profiles, provide an ecologically meaningful graphical representation of interaction niches, and explore the effect of latent traits on network structure.

ecology↗

bioRxiv2024

Zero-shot animal behavior classification with image-text foundation models

1. Understanding the behavior of animals in their natural habitats is critical to ecology and conservation. Camera traps are a powerful tool to collect such data with minimal disturbance. They however produce very a large quantity of images, which can make human-based annotation cumbersome or even impossible. While automated species identification with artificial intelligence has made impressive progress, automatic classification of animal behaviors in camera trap images remains a developing field. 2. Here, we explore the potential of foundation models, specifically Vision Language Models (VLMs), to perform this task without the need to first train a model, which would require some level of human-based annotation. Using an original dataset of alpine fauna with behaviors annotated by participatory science, we investigate the zero-shot capabilities of different kind of recent VLMs to predict behaviors and estimate behavior-specific diel activity patterns in three ungulate species. 3. Our results show that using these methods, it is possible to achieve accuracies over 91% in behavior classification and produce activity patterns that closely align with those derived from participatory science data (overlap indexes between 84% and 90%). 4. These findings demonstrate the potential of foundation models and vision-language models in ecological research. Ecologists are encouraged to adopt these new methods and leverage their full capabilities to facilitate ecological studies.

ecology↗

bioRxiv2023

Abundance of mega-carcasses does not profoundly affect trends in the population dynamics of a facultative scavenger predator and its prey

The interplay between facultative scavenging and predation has gained interest in the last decade. The prevalence of scavenging induced by the availability of large carcasses may modify predator density or behaviour, potentially affecting prey. In contrast to behavioural mechanisms through which scavenging affects predation, the demographic effects of facultative scavenging on predator and prey populations remain poorly studied. We used the semi-natural experimental opportunity in Hwange National Park, Zimbabwe, where contrasted management measures (culling and artificial supply of water) have led to fluctuations in elephant carrion abundance, to identify the consequences of facultative scavenging on the population dynamics of a large mammalian carnivore, the spotted hyaena (Crocuta crocuta), and its prey. Using a 50-year dataset and Multivariate Autoregressive State Space models, we estimated hyaena and prey densities over four periods contrasted in elephant carrion availability due to management practices. Models that allow hyaena and their prey populations growth rate to vary depending on these four periods contributed significantly to explain variations in their density, which is consistent with an effect of management measures on the population dynamics of hyaena and its prey. Although our results support a predominant role of bottom-up mechanisms, whereby hyaena density is driven by herbivore density, itself driven by resources availability, some subtle patterns of densities could be interpreted as consequences of changes in predation pressure following changes in scavenging opportunities. We discuss why signals of prey and predator population dynamics decoupling are less likely to be observed in systems with a high diversity of prey, such as African savannas, and why inputs of mega-carcasses as pulsed resources hardly impacted top-down relationships in the long run. This study represents a first investigation of the long-term effects of carrion pulses, whose frequency may increase with climate changes, on the classical predator-prey coupling for large mammals.

ecology↗

bioRxiv2023

Beyond accuracy : score calibration in deep learning models for camera trap image sequences

In ecological studies, machine learning models are increasingly being used for the automatic processing of camera trap images. Although this automation facilitates and accelerates the identification step, the results of these models may lack interpretability and their immediate applicability to ecological downstream tasks (e.g occupancy estimation) remain questionable. In particular, little is known about their calibration, a property that guarantees that confidence scores can be reliably interpreted as probabilities that a models predictions are true. Using a large and diverse European camera trap dataset, we investigate whether deep learning models for species classification in camera trap images are well calibrated, or in contrast over/under-confident. Additionally, as camera traps are often configured to take multiple photos of the same event, we also explore the calibration of predictions at the sequence level. Finally, we study the effect and the practicality of a post-hoc calibration method, i.e. temperature scaling, for predictions made at image and sequence levels. Based on five established models and three independent test sets, our findings show that, using the right methodology, it is possible to enhance the interpretability of the confidence scores, with clear implication for, for instance, the calculation of error rates or the selection of confidence score thresholds in ecological studies making use of artificial intelligence models.

ecology↗

bioRxiv2023

PhylteR: efficient identification of outlier sequences in phylogenomic datasets

In phylogenomics, incongruences between gene trees, resulting from both artifactual and biological reasons, can decrease the signal-to-noise ratio and complicate species tree inference. The amount of data handled today in classical phylogenomic analyses precludes manual error detection and removal. However, a simple and efficient way to automate the identification of outliers from a collection of gene trees is still missing. Here, we present PhylteR, a method that allows a rapid and accurate detection of outlier sequences in phylogenomic datasets, i.e. species from individual gene trees that do not follow the general trend. PhylteR relies on DISTATIS, an extension of multidimensional scaling to 3 dimensions to compare multiple distance matrices at once. In PhylteR, these distance matrices extracted from individual gene phylogenies represent evolutionary distances between species according to each gene. On simulated datasets, we show that PhylteR identifies outliers with more sensitivity and precision than a comparable existing method. We also show that PhylteR is not sensitive to ILS-induced incongruences, which is a desirable feature. On a biological dataset of 14,463 genes for 53 species previously assembled for Carnivora phylogenomics, we show (i) that PhylteR identifies as outliers sequences that can be considered as such by other means, and (ii) that the removal of these sequences improves the concordance between the gene trees and the species tree. Thanks to the generation of numerous graphical outputs, PhylteR also allows for the rapid and easy visual characterisation of the dataset at hand, thus aiding in the precise identification of errors. PhylteR is distributed as an R package on CRAN and as containerized versions (docker and singularity).

evolutionary biology↗

bioRxiv2022

Mycobacterium tuberculosis genetic features associated with pulmonary tuberculosis severity

Mycobacterium tuberculosis (Mtb) infections result in a wide spectrum of clinical presentations but without proven Mtb genetic determinants. Herein, 234 pulmonary tuberculosis (TB) patients were stratified according to TB disease severity and Mtb genetic features were explored using whole genome sequencing, including heterologous single nucleotide polymorphism (SNP) calling to explore micro-diversity. Clinical isolates from patients with mild TB carried mutations in genes associated with host-pathogen interaction, while those from patients with moderate/severe TB carried mutations associated with regulatory mechanisms. Genome-wide association study identified a SNP in the promoter of the gene coding for the virulence regulator EspR associated with moderate/severe disease. Structural equation modelling and model comparisons indicated that TB severity was associated with the detection of Mtb micro-diversity within clinical isolates and to the espR SNP. Taken together, these results provide a new insight to better understand TB pathophysiology and could provide new prognosis tool for pulmonary TB severity.

microbiology↗

bioRxiv2021

Within-host genetic micro-diversity of Mycobacterium tuberculosis and the link with tuberculosis disease features

Tuberculosis (TB), caused by Mycobacterium tuberculosis (Mtb) complex, is still the number one deadly contagious disease. Mtb infection results in a wide spectrum of clinical presentations and severity symptoms, but without proven Mtb genetic determinants. Thanks to a collection of 355 clinical isolates with associated patients clinical data, we showed that Mtb micro-diversity within patient isolates is strongly correlated with TB-associated severity scores. Interestingly, this diversity is driven by a selection pressure to adapt to different lifestyles related to the infection site. Taken together, these results provide a new insight to better understand TB pathophysiology. Furthermore, Mtb micro-diversity could be envisioned as a new prognostic tool to improve the management of TB patients.

microbiology↗

bioRxiv2020

Common causes drive negative correlations between nuclear genetic and species level biodiversity

The processes that give rise to species richness gradients are not well understood, but may be linked to resource-based limits on the number of species a region can support. Ecological limits placed on regional species richness would also limit population sizes, suggesting that these processes could also generate genetic diversity gradients. If true, we might better understand how broad-scale biodiversity patterns are formed by identifying the common causes of genetic diversity and species richness. We develop a hypothetical framework based on the consequences of regional variation in ecological limits to simultaneously explain spatial patterns of species richness and neutral genetic diversity. Repurposing raw genotypic data spanning 38 mammal species sampled across 801 sites in North America, we show that estimates of genome-wide genetic diversity and species richness share spatial structure. Notably, species richness hotspots tend to harbor lower levels of within-species genetic variation. A structural equation model encompassing eco-evolutionary processes related to resource availability, habitat heterogeneity, and human disturbance explained 78% of variation in genetic diversity and 74% of the variation in species richness. These results suggest we can infer broad-scale patterns of species and genetic diversity using two simple environmental measures of resource availability and ecological opportunity.

evolutionary biology↗