Belgium

Supervised machine learning is widely used to detect fine-scale animal behaviour from accelerometer data. However, there is currently no standardised protocol for verifying the accuracy of this technique, and the risks associated with common machine learning verification pitfalls have yet to be widely recognised. One such pitfall is model overfitting, which occurs when a model becomes too closely aligned with the training data, memorising noise rather than capturing generalisable patterns. As a result, while the model may show high performance on the training set, its ability to generalise to new, unseen data is compromised. This limitation can be detected with the application of robust time-series specific validation techniques. We conducted a systematic review of 119 accelerometer-based behaviour classification supervised machine learning papers to see what proportion of existing literature conducted appropriate time-series validation. We found that 79% (94 papers) did not validate their models sufficiently well to identify potential overfitting. Although this does not inherently imply that these models are overfit, the absence of independent test sets limits the interpretability of their results. To address these challenges, we provide a theoretical overview of overfitting in the context of time-series accelerometer data and propose succinct, ecologist-accessible guidelines for optimal time-series validation techniques. Our aim is to equip ecologists with the tools necessary to adapt general machine learning theory to the specific requirements of time-series biologging, facilitating reliable overfitting detection and continuing to advance the field.

SEB Conference 2025

Practical Guidelines for Supervised Machine Learning Model Validation in Accelerometer-based Animal Behaviour Classification

accelerometer

over-fitting

validation

### Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

Message from the General Chair: 
*It is my great pleasure and honor to welcome you to the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025), held in beautiful Vienna, Austria, from July 27 to August 1, 2025. ACL2025continues our field’s tradition of excellence in scholarship, innovation, and inclusivity, and I am deeply grateful to the many volunteers who have worked tirelessly to bring this event to life.* 
[Read more](https://drive.google.com/file/d/1GI_hvOpjswAuYdUTromfeDiPpCcqidwg/view?usp=sharing)

To access this event page, you need to log in with the **email address you registered with**. Access credentials will be sent to your email from Underline - subject line "Welcome to ACL 2025". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

To access this event page, you need to log in with the **email address you registered with**. 

ACL 2025

Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

poster

You need to log in with the email address you registered with. 

Please log in to explore this event.

If you see this page it means you are not registered. You can register via [this link](https://www.sebiology.org/events/seb-annual-conference-antwerp-2025/registration.html). 

If you are registered and feel like you are seeing this message by mistake, please make sure you are logged in with the same email that you registered with. 

Please register!

SEB’s Annual Conference is celebrated for creating the perfect composition of research, new discoveries and collaborative connections. Be creative and innovative by exchanging ideas and knowledge with masters of biology from all over the world.

Background: There is individual variability in the response of core body temperature (CBT) during exposure to heat load. Ewes classified as having high rectal temperatures on hot days before breeding had smaller lambs, with those lambs having lower survival rate than the lambs from ewes that had a lower rectal temperature on hot days before breeding. The ability to classify animals as having either a high or low body temperature suggests the possibility that a thermal phenotype may exist in CBT.
Aims: Rumen temperature (Trumen ) was recorded as a proxy for CBT in ewes during summer to determine whether a repeatable thermal phenotype could be identified within individuals.
Methods: Trumen  was recorded every five minutes over 34 days during the hottest period of summer in 43 ewes kept in a paddock with no access to shade, while data from a weather station were used to calculate the average daily temperature-humidity index. On each day, each ewe was ranked based on its 95th  percentile Trumen  for that day. The consistency of the ranking was assessed across days by correlating each ewe’s ranking on a given day with its ranking on every other day during summer.
 Results: The individual ranking of Trumen  was stable over time, with the correlation coefficient between days ranging between 0.25 and 0.75.
 Conclusion and Implications: Ewes have a thermal phenotype in CBT during summer. This thermal phenotype may impact animal performance and welfare.

Identifying a thermal phenotype from field measurement of the core temperature of ewes

The frequency of climate change-induced hypoxic events defined as a decrease in dissolved oxygen levels in water is increasing at an unprecedented rate. Enclosed ecosystems are particularly vulnerable, as organisms in these environments cannot escape hypoxia and must develop ways to cope with its chronic effects to avoid extinction. When exposed to short-term hypoxia, fish can rapidly adjust their phenotype through direct plasticity. If environmental conditions remain similar across generations, intergenerational plasticity can also occur. When the stressor becomes persistent, adaptation may take place. Therefore, more research is needed to understand how fish respond to multigenerational hypoxia. To investigate this, we studied two populations of Gasterosteus aculeatus collected from separate streams, one that has been exposed to hypoxia for a decade and one that has not. Within each population, we bred two generations, exposing offspring to either normoxia (100% dissolved oxygen, DO) or fluctuating hypoxia (30% DO at night and 100% DO during the day). This experimental setup resulted in eight groups, allowing us to examine direct plasticity, intergenerational plasticity (when offspring matched parental environmental conditions), and adaptation between the populations. We measured various fitness-related traits, including social and risk-taking behavior, metabolic rate (standard metabolic rate, SMR), and hypoxia tolerance. When directly exposed to hypoxia, fish reduced their social and risk-taking behaviors but showed no change in metabolic rate. However, after multiple generations of exposure, fish maintained active social and risk-taking behaviors while decreasing their SMR. This study disentangles the responses to hypoxia within and across generations used by fish.

Resilience in a hypoxic world: Behavioral and metabolic responses of stickleback populations exposed to multigenerational fluctuating hypoxia

In a previous study, it was: shown that this species had a maximum metabolic rate (MMR) during a critical thermal maximum while swimming (CTSMax ) test that was 60% higher than measured during a critical thermal maximum (CTMax ) test (i.e., on resting fish); and suggested that this higher MMR allowed these swimming fish to tolerate only slightly lower maximum temperatures (37.5 vs. 38.9o C, respectively). However, the mechanism(s) mediating this difference in MMR was/were not known. Thus, we performed CTMax  and CTSMax  tests in a swim-tunnel respirometer on snapper that had been implanted with Transonic®  blood flow probes so that heart rate (fH ), stroke volume (SV ), cardiac output (Q̇) and oxygen extraction (ṀO2  / Q̇) could be measured. The ~ 1.75 fold higher MMR in schoolmaster snapper given a CTSMax  vs. CTMax  test was not due to a greater contribution of SV  to Q̇ in swimming fish exposed to acute warming as was previously hypothesized. Instead the increase in fH  was ~ 40 beat min-1  (25% higher) in fish given the CTSMax  test, and ṀO2  / Q̇ was ~ 1.55-fold greater. This increase in ṀO2  / Q̇ may be related to improved oxygen uptake at the gills (and thus oxygen content of the arterial blood) and/or increased extraction efficiency in swimming fish at warm temperatures. Nonetheless, these data are another example of the importance of ṀO2  / Q̇ in determining the tolerance of fish exposed to combined stressors, and how estimating thermal tolerance based on measures of fH  alone may be misleading.

Increased Oxygen Extraction is Key to the Ability of Schoolmaster Snapper (Lutjanus apodus) to Swim Maximally at High Temperatures

Burkholderia cepacia is a bacterium commonly found in humid environments and water sources, which can cause severe respiratory infections in immunocompromised patients or those with specific chronic diseases. Due to its unique survival and proliferation abilities, this bacterium can thrive in natural water bodies, specific sterile or non-sterile preparations, and in environments where antibiotics and disinfectants are present and in nutrient-deficient conditions. This makes it a significant threat to the water supply systems in medical facilities. Many broad-spectrum antibiotics, such as penicillins and polymyxins, are ineffective against B. cepacia, complicating the treatment process, and its transmission characteristics increase the likelihood of healthcare-associated infections. Therefore, conducting in-depth analyses of this bacterium is crucial to propose more precise and effective prevention and treatment strategies. This study collected approximately 80 strains isolates of B. cepacia. First, random amplified polymorphic DNA (RAPD) typing was performed, and then their potential to cause nosocomial infections was analyzed. Subsequently, drug susceptibility testing was conducted using the disk diffusion method to identify effective antibiotics, aiming to prevent the potential spread of this bacterium.

Characteristics and drug susceptibility analysis of Burkholderia cepacia isolates

With advanced information technology geared toward biochemical assays, molecular diagnostics, and physiological measurements across the animal kingdom, animal biological research nevertheless suffers from a more archaic laboratory workflow, involving fragmented data acquisition, manual sample handling, and spurious quality control, leading to high variability and turnaround time. Complementary to the challenges, we developed intelligent AI-based automation in life sciences for improved data acquisition, processing, and analysis, unlike anything, thus far increasing efficiency and resource management. The system allows optimizing the workflow with machine learning algorithms that continuously tune the processing parameters based on the samples' performances to improve accuracy while minimizing error. A high-throughput data management framework allows real-time synchronization of biochemical, molecular, and physiological data sets. Intelligent quality control ensures anomaly detection and thus also increases the reliability of data and reproducibility across experiments. When applied to animal physiological and biochemical research, a reduction in turnaround time of the order of 18% to manual labor of 30% was described. Energy consumption was minimized by 30% with the use of automated processes and 180 kg of laboratory waste was greatly reduced, therefore providing such green benefits for high-throughput studies. Further work will expand the biomonitoring driven by AI on wildlife physiology and the predictive modeling of metabolic and ecological research, giving a more scalable framework to integrate AI-driven automation into animal biology and research to improve production efficiency, reproducibility, and sustainability moving forward.

AI-Driven Automation in Animal Biological Research: Enhancing Data Processing and Resource Efficiency

Burkholderia cepacia complex is a Gram-negative, non-fermentative bacterium in humid environments like soil, water, and plants. Initially isolated from onion bulbs, it is now detected in freshwater, agricultural runoff, and various ecological niches. Its adaptability allows it to thrive in natural and artificial aquatic systems, posing potential risks to animals and ecosystems. Amid growing concerns about antibiotic resistance, bacteriophages present a promising alternative for bacterial control. This study aims to isolate phages infecting Burkholderia cepacia complex from various freshwater sources and describe their characteristics. Seventy-eight strains were collected and genotyped using random amplified polymorphic DNA (RAPD) analysis. Phage screening on these isolates identified several effective lytic phages for potential bacterial control applications. By exploring phage-bacterium interactions in natural settings, this research proposes novel biocontrol strategies to inhibit the proliferation of Burkholderia cepacia complex in water sources, potentially offering eco-friendly solutions for mitigating bacterial contamination in animal habitats, agriculture, and aquaculture.

Isolation and Characterization of Bacteriophages Targeting Burkholderia cepacia Complex

Climate change disrupts aquaculture systems increasingly in terms of water temperature, pH, and microbial community structure, thereby deteriorating water quality and increasing disease risk. While subtropical islands in East Asia undergo drastic seasonal temperature variation and extreme weather events, they also further influence the aquaculture environment. To that end, it is essential to recognize these shifts in microbes in formulationally sustainable fish culture. The research has conducted investigations on the microbial community structure in the aquaculture ponds on a subtropical island, emphasising the relationship between climate-induced changes in water quality. We shall identify critical environmental circumstances affecting microbial population dynamics, including their roles in aquaculture sustainability. Water samples were collected from aquaculture ponds across central and southern regions, targeting economically important species such as milkfish and shrimp. Microbial communities were analysed through 16S rRNA sequencing, and physicochemical parameters, including temperature, dissolved oxygen, pH, and nitrogen compounds, were monitored. The microbial shifts were shown to be highly seasonal, with disruptions of the nitrogen-cycler bacteria with a strong effect on water quality. This advance underscores the importance of monitoring microbial communities as a means of acclimatizing to climate hazards within the aquaculture sector, and thus should be in place to formulate an effective water quality regulation strategy and promote a microbial regime that grants fish health and production efficiency: a prerequisite for the resilience and sustainability of aquaculture industries in East Asia against environmental change.

Microbial community dynamics in aquaculture under climate change: impacts on water quality

Lactose intolerance presents high incidence, although the mechanisms have not yet been fully elucidated and there are few/limited experimental models for their study. Some studies suggest a sexually dimorphic effect and a greater predisposition to psychiatric disorders. The objectives of this study were to propose experimental models of lactose intolerance, evaluating the manifestation of behaviours related to psychiatric disorders and the incidence in each sex. Moreover, it was objective to elucidate the neurobiological processes involved. Male and female adult Wistar rats were exposed for 14 consecutive days to two types of diets rich in lactose: a solution containing 14% lactose and a commercial formula of milk. Clinical signs were monitored: body weight and stool consistency. Behaviours related to psychiatric disorders were evaluated: light-dark box (anxiety) and social preference tests, as well as brain dopamine levels. Both diets impaired the stool consistency, with a peak effect on the 12th day of exposure. From there, the results were sexually dimorphic. Males presented anxiolytic behaviour after the lactose diet. The socialization of males was impaired by both diets, especially after the commercial formula. The behavioural impairments found in males were attributed to the modulating role of lactose on the dopaminergic system. In females, the lactose-rich diets reduced cortical dopamine levels, which resulted in an anxiogenic effect. In conclusion, both lactose-rich diets (14% lactose and commercial formula) induced clinical signs of lactose intolerance in male and female rats, accompanied by sexually dimorphic behavioural impairments associated with psychiatric disorders that were explained by dopaminergic modulation.

Study of experimental models of lactose intolerance in rats: sexual dimorphism and relationship with psychiatric disorders

Cold-water fishes, such as salmonids, face environmental stresses resulting from climate change. We examined the thermal acclimation response of the hatchery-reared brook trout (Salvelinus fontinalis) to explore how the metabolic function and contractile properties of their myotomal muscle might shift to mitigate the impact of a warming environment. When tested at a common temperature (10o C), metabolic rate in fish acclimated to 20o C was significantly lower than those acclimated to 10o C. Contractile properties such maximum muscle shortening velocity and isometric relaxation showed corresponding changes, with warm acclimated fish showing slower contraction kinetics. These fish also showed changes in the expression of a wide variety of muscle and metabolic genes with elevated temperature, thus providing a mechanistic foundation for their thermal acclimation. We carried out a parallel study on wild brook trout collected in summer, fall and winter to explore if a native population of these fish undergoes the same thermal acclimation response observed with our hatchery fish.  These fish showed similar but not identical shifts in muscle function and gene expression with a warming environment, suggesting that wild populations of brook trout may vary in their thermal acclimation response across the range of this cold-water salmonid.

Metabolic and Contractile Shifts in Brook Trout Myotomal Muscle Acclimated to Warm Water

Calcification is a widespread process in marine organisms and depends on tight control of pH and carbonate chemistry. Sea urchin larvae have been extensively used to study the transport mechanisms underlying the formation of an elaborate calcitic endoskeleton. The generation of this complex mineral structure is initiated in primary mesenchyme cells (PMCs), where amorphous calcium carbonate (ACC) is formed in intracellular vesicles. The intracellular precipitation of ACC requires a specific chemical environment in which pH plays a critical role. However, the pH dynamics within these vesicles and their role in carbon transport are still poorly understood. We developed in vivo methods to characterize vesicular pH in PMCs of Stongylocentrotus purpuratuslarvae to unravel the underlying mechanisms of proton and carbon transport. Inverted confocal microscopy in combination with the ratiometric fluorescent dye SNARF-1 allowed us to accurately measure the pH of intracellular vesicles in PMCs. Simultaneous tracking of endocytosed calcein accompanied by pH recordings allowed a high spatial and temporal resolution of pH dynamics in Ca2+  rich vesicles. This system provides a unique model for live- tissue imaging and sheds light on pH regulatory mechanisms in intracellular calcifying vesicles. Preliminary results showed the presence of acidic and alkaline vesicle populations in PMCs, which can regulate their luminal pH to maintain a favorable chemical environment forCaCO3  formation. Future studies are planned to understand how the larvae can cope with environmental stressors such as ocean acidification, where the decrease in seawater pH is accompanied by a decrease in carbonate ion concentration.

Premium content

Next from SEB Conference 2025

Identifying a thermal phenotype from field measurement of the core temperature of ewes

Similar lecture

LETS-C: Leveraging Text Embedding for Time Series Classification