Biology subjects

Mills, S. C.

Publications and source records attributed to Mills, S. C..

6 recordsLinked to original sources

bioRxiv2026

Contrasting defensive strategies underlie differential susceptibility of corals to crown-of-thorns sea star (CoTS; Acanthaster cf. solaris) predation

Crown-of-thorns sea star (CoTS), Acanthaster cf. solaris, outbreaks are a major cause of hard coral cover decline across the west Pacific, threatening coral reefs. Coral taxa vary in susceptibility to CoTS predation from preferred (Acropora spp.) to non-preferred (Porites spp.), yet the mechanisms underlying these differences are poorly understood. We investigated coral defenses during an ongoing CoTS outbreak in Moorea, French Polynesia by examining gene expression (including putative toxin genes) in healthy and actively predated colonies of a preferred (Acropora hyacinthus) and a non-preferred (Porites sp.) coral prey species. During predation, A. hyacinthus exhibited molecular signatures of cellular stress responses involving oxidative stress signalling, inflammation, and tissue proteolysis. In contrast, Porites sp. showed enrichment of genes involved in mitochondrial metabolic adjustment and aerobic metabolism, suggesting metabolic compensation to maintain cellular function. Furthermore, A. hyacinthus demonstrated a reactive defense behaviour by differentially expressing toxins (e.g., kunitz-type neurotoxins) while Porites sp. employed constitutive expression of all putative toxins regardless of active predation, suggesting a proactive defense strategy. Together, these findings suggest that preferred and non-preferred coral prey exhibit fundamentally different molecular and defensive strategies during CoTS predation, shedding light on the evolutionary arms race between corals and their predators.

molecular biology↗

bioRxiv2025

Coral venom and toxins as protection against crown-of-thorns sea star attack

Crown-of-thorns sea star (CoTS) outbreaks are one of the leading causes of hard coral cover decline across the Indo-Pacific, posing a major threat to the health and resilience of coral reefs. However, the drivers underlying feeding on preferred (e.g., Acropora spp.) versus non-preferred (e.g., Porites spp.) are poorly understood. We hypothesised that coral venom may influence CoTS food preferences. We investigated whether coral venom toxin and peptide families may drive CoTS prey preferences by comparing the genomes and transcriptomes of preferred (five Acropora species) and non-preferred (five Porites species and Echinopora lamellosa) prey species of CoTS. We constructed databases of known cnidarian venom toxins, and along with the full UniProtKB/Swiss-Prot Tox-Prot database, used these to identify toxin peptides and investigate function and phylogeny. The most abundant toxins across all coral species included kunitz-type neurotoxins, neurotoxic turripeptides, snake venom lectins, toxic proteases and actinoporins. There were proteins present only in certain Porites species but completely absent from all Acropora species (e.g., tereporin/conoporin, snake venom peptides) and vice versa (e.g., sarafotoxin). Further, Porites species contained a homolog to conkunitzin, a toxin known to disintegrate the tube feet of CoTS, suggesting a potential mechanism for their lower susceptibility to predation. We also observed a greater diversity of jellyfish-like proteins in CoTS-exposed Porites species compared to naive Porites species, suggesting these proteins deter CoTS. These findings have direct applications to assessing reef corals susceptibility to future CoTS outbreaks and active reef management.

bioinformatics↗

bioRxiv2024

Unlocking the soundscape of coral reefs with artificial intelligence

Passive acoustic monitoring can offer insights into the state of coral reef ecosystems at low-costs and over extended temporal periods. Comparison of whole soundscape properties can rapidly deliver broad insights from acoustic data, in contrast to the more detailed but time-consuming analysis of individual bioacoustic signals. However, a lack of effective automated analysis for whole soundscape data has impeded progress in this field. Here, we show that machine learning (ML) can be used to unlock greater insights from reef soundscapes. We showcase this on a diverse set of tasks using three biogeographically independent datasets, each containing fish community, coral cover or depth zone classes. We show supervised learning can be used to train models that can identify ecological classes and individual sites from whole soundscapes. However, we report unsupervised clustering achieves this whilst providing a more detailed understanding of ecological and site groupings within soundscape data. We also compare three different approaches for extracting feature embeddings from soundscape recordings for input into ML algorithms: acoustic indices commonly used by soundscape ecologists, a pretrained convolutional neural network (P-CNN) trained on 5.2m hrs of YouTube audio and a CNN trained on individual datasets (T-CNN). Although the T-CNN performs marginally better across the datasets, we reveal that the P-CNN is a powerful tool for identifying marine soundscape ecologists due to its strong performance, low computational cost and significantly improved performance over acoustic indices. Our findings have implications for soundscape ecology in any habitat. Author SummaryArtificial intelligence has the potential to revolutionise bioacoustic monitoring of coral reefs. So far, a limited set of work has used machine learning to train detectors for specific sounds such as individual fish species. However, building detectors is a time-consuming process that involves manually annotating large amounts of audio followed by complicated model training, this must then be repeated all over again for any new dataset. Instead, we explore machine learning techniques for whole soundscape analysis, which compares the acoustic properties of raw recordings from the entire habitat. We identify multiple machine learning methods for whole soundscape analysis and rigorously test these using datasets from Indonesia, Australia and French Polynesia. Our key findings show use of a neural network pretrained on 5.2m hours of unrelated YouTube audio offers a powerful tool to produce compressed representations of reef audio data, conserving the datas key properties whilst being executable on a standard personal laptop. These representations can then be used to explore patterns in reef soundscapes using "unsupervised machine learning", which is effective at grouping similar recordings periods together and dissimilar periods apart. We show these groupings hold relationships with ground truth ecological data, including coral coverage, the fish community and depth.

ecology↗

bioRxiv2023

Introducing flocker: an R package for flexible occupancy modeling via brms and Stan

O_LIOccupancy models are a widespread tool for analyzing biological survey data, but packages for fitting these models offer a limited variety of effects structures. C_LIO_LIWe developed Rpackage flocker to connect occupancy models (single- and multi- species; single- and multi-season) to the uniquely powerful formula syntax of Rpackage brms. C_LIO_LIUsing familiar formula-based syntax, flocker models can readily incorporate a wide array of effects structures, including phylogenetic random effects, splines, Gaussian processes, autoregressive structures, monotonic effects, and nonlinear predictors. These are available for use in formulas for occupancy, detection, colonization, extinction, and autologistic terms (as applicable to the model type). flocker additionally provides functionality for data simulation, posterior prediction, and model comparison, following well-documented statistical decisions that we put forward as best practices. C_LIO_LIWe anticipate that flocker will facilitate the work of practitioners who seek added realism in occupancy models. We further hope that flockers synthesis of these models will help inform best practices around occupancy modeling. C_LI

ecology↗

bioRxiv2022

Validation of a novel fully immersive Virtual Reality setup with a behavioural study of freely-moving fish

Virtual Reality (VR) enables standardised stimuli to invoke behavioural responses in animals, however, in fish studies VR has been limited to either basic virtual stimulation projected below the bowl for freely-swimming individuals or a simple virtual arena rendered over a large field-of-view for head-restrained individuals. We developed a novel immersive VR setup with real-time rendering of animated 3D scenarios, validated in a proof of concept study on the behaviour of coral reef post-larval fish. Fish use a variety of cues to select a habitat during the recruitment stage, and to recognize conspecifics and predators, but which visual cues are used remains unknown. We measured behavioural responses of groups of five convict surgeonfish (Acanthurus triostegus) to simulations of habitats, static or moving shoals of conspecifics, predators, and non-aggressive heterospecifics. Post-larval fish were consistently attracted to virtual corals and conspecifics presented statically, but repulsed by their predators (bluefin jacks, Caranx melampygus). When simulated shoals passed nearby repeatedly, they were again attracted by conspecifics showing a tendency to follow the shoal, whereas they moved repeatedly to the back of the passing predator shoal. They also discriminated between species of similar sizes: they were attracted more to conspecifics than butterflyfish (Forcipiger longirostris), and repulsed more by predators than parrotfish (Scarus psittacus). The quality of visual simulations was high enough to identify between visual cues - size, body shape, colour pattern - used by post-larval fish in species recognition. Despite a tracking technology limited to fish 2D positions in the aquarium, preventing the real-time updating of the rendered viewpoint, we could show that VR and modern tracking technologies offer new possibilities to investigate fish behaviour through the quantitative analysis of their physical reactions to highly-controlled scenarios.

animal behavior and cognition↗

bioRxiv2021

Biogeographic multi-species occupancy models for large-scale survey data

O_LIEcologists often seek to infer patterns of species occurrence or community structure from survey data. Hierarchical models, including multi-species occupancy models (MSOMs), can improve inference by pooling information across multiple species via random effects. Originally developed for local-scale survey data, MSOMs are increasingly applied to larger spatial scales that transcend major abiotic gradients and dispersal barriers. At biogeographic scales, the benefits of partial pooling in MSOMs trade off against the difficulty of incorporating sufficiently complex spatial effects to account for biogeographic variation in occupancy across multiple species simultaneously. C_LIO_LIWe show how this challenge can be overcome by incorporating pre-existing range information into MSOMs, yielding a biogeographic multi-species occupancy model (bMSOM). We illustrate the bMSOM using two published datasets: Parulid warblers in the United States Breeding Bird Survey, and entire avian communities in forests and pastures of Colombias West Andes. C_LIO_LICompared to traditional MSOMs, the bMSOM provides dramatically better predictive performance at lower computational cost. The bMSOM avoids severe spatial biases in predictions of the traditional MSOM and provides principled species-specific inference even for never-observed species. C_LIO_LIIncorporating pre-existing range data enables principled partial pooling of information across species in large-scale MSOMs. Our biogeographic framework for multi-species modeling should be broadly applicable in hierarchical models that predict species occurrences, whether or not false-absences are modeled in an occupancy framework. C_LI

ecology↗