Royal Society Open Science

Dental morphology and wear patterns provide insight into the dietary adaptations and ecological niches of living and extinct herbivores. Traditional classification statistics such as Linear Discriminant Analysis (LDA) are limited by assumptions of linearity, normality, and homoscedasticity. This study quantifies mesowear, the shape of molar cusps resulting from occlusal wear, and evaluates the performance of non-linear machine learning models in predicting herbivore diets based on geometric morphometric (GMM) data from adult mandibular second molars (M2) in bovids. We applied Generalized Procrustes Analysis and Principal Component Analysis (PCA) to digitized occlusal shape coordinates from 132 M2 specimens across 64 species. Using the resulting principal component scores, we compared the classification accuracy of LDA with three non-linear models: Random Forest, K-Nearest Neighbors, and Gradient Boosting. While LDA achieved a cross-validated accuracy of just 31%, all non-linear models achieved 99% cross-validation accuracy and 90% test accuracy, demonstrating substantially improved performance. Misclassification analyses revealed that non-linear models more effectively captured complex shape differences, particularly among species with overlapping wear patterns. Our findings support the integration of machine learning with geometric morphometrics to quantify mesowear and improve dietary classification, providing a framework for robust paleoecological inference.

Advanced footwear technology (AFT) shoes incorporate increased sole thickness and compliant midsole materials that may alter running biomechanics. While these effects have been widely studied during level running, little is known about how sole thickness influences running style and stability during uphill running. This study examined the effects of two AFT shoes differing in sole thickness (35 mm-AFT35; 50 mm-AFT50) and a traditional control shoe (27 mm-CON27) on running style and stability during uphill running. Seventeen experienced male runners performed treadmill running at a 10% incline at 6.5 and 10 km/h in three shoe conditions. Running style was assessed using duty factor, normalized step frequency, center-of-mass oscillation, vertical and leg stiffness, and lower-limb joint kinematics. Running stability was evaluated using local dynamic stability via the maximum Lyapunov exponent and detrended fluctuation analysis of stride time. Duty factor and normalized step frequency did not differ between shoes. However, AFT shoes showed greater center-of-mass oscillation (p = 0.004), lower vertical stiffness (p = 0.022) compared to CON27. Joint kinematics revealed significant shoe effects at the ankle (p = 0.001), particularly increased dorsiflexion and eversion in AFT conditions. Running stability showed only minor changes. Local dynamic stability differed at the trunk (p = 0.027), with reduced stability in AFT50 compared with CON27 (p = 0.006), while global stability remained unchanged. No shoe x speed interactions were observed for any variable. Overall, uphill running style and stability remained largely preserved across shoe conditions, suggesting that sole thickness alone had limited influence.

O_LICardiac rate and rhythm reveal how animals adapt physiologically to day-to-day challenges, with consequences for health and fitness. However, these data remain difficult to collect in wild animals, despite their relevance for individual health and fitness. C_LIO_LIHere, we present a system for collecting and transmitting long-term, fine-scaled physiological data in wild animals. We implanted Bluetooth-enabled cardiac and physiological monitor devices in three wild adult female baboons in the Amboseli ecosystem in Kenya and paired these devices with collars that enabled remote data downloads over long-range wide area network (LoRaWAN). C_LIO_LIThe system performed well over >10 months, providing the first long-term cardiac data in wild primates. The baboons showed strong circadian patterns in heart rate, heart rate variability, and activity. We also present data on one female who left her social group for unknown reasons; while alone she exhibited higher heart rate variability, lower activity, and evidence of disrupted sleep. C_LIO_LIIn sum, physiologgers paired with low-energy methods of remote data retrieval are powerful tools for investigating physiology in wild animals on timescales that extend over many months, with minimal disruption to their behavior. C_LI

Widow spiders of the genus Latrodectus are important animals for biomedical, pest and conservation research. Here, we present the assembled genomes of two closely related Latrodectus species: the Australian L. hasselti and the New Zealand endemic L. katipo. The genome of L. katipo consists of 13 scaffolds likely corresponding to chromosomes (90% of the total length) and 1267 short scaffolds (10%). It has a total length of 1.5 Gbp and BUSCO of 94.9%. The genome of L. hasselti consists of 379 scaffolds and has a total length of 1.7 Gbp and a BUSCO score of 95.4%. The repeat content is very similar in both genomes with a total proportion of 37.2% for L. katipo and 39.9% for L. hasselti. Genome annotation predicted 12706 and 15111 genes for L. katipo and L. hasselti respectively. An ortholog analysis shows large overlap between orthogroups suggesting either duplication events in L. hasselti or loss of genes in L. katipo.

Background: Radiographic detection of caries lesions adjacent to restorations is challenging due to limitations of two-dimensional imaging and difficulties distinguishing true lesions from restorative or anatomical radiolucencies. Artificial intelligence (AI)-based clinical decision support systems (CDSSs) have been introduced to assist radiographic interpretation; however, different AI tools may yield variable diagnostic outputs, and their comparative performance remains unclear. Objective: To compare the diagnostic performance of commercial and experimental AI algorithms for detecting secondary caries lesions on bitewings. Methods: This cross-sectional diagnostic accuracy study included 200 anonymized bitewings comprising 885 restored tooth surfaces. A consensus group reference standard identified all surfaces with a caries lesion and classified each lesion by type (primary/secondary) and depth (enamel-only/dentin-involved). Five commercial (Second Opinion, CranioCatch, Diagnocat, DIO Inteligencia, and Align X-ray Insights) and three experimental (Mask R-CNN-based and Mask DINO-based) systems were tested. Diagnostic performance was expressed through sensitivity, specificity, and overall accuracy (95% CI). Comparisons used generalized estimating equations, adjusted for clustered data. Results: Specificity was high across all systems (0.957-0.986), confirming accurate recognition of non-carious surfaces, whereas sensitivity was moderate (0.327-0.487), reflecting frequent missed detections of enamel and dentin lesions. Accuracy ranged from 0.882 to 0.917, with no significant differences among models (p >= 0.05). Confounding factors, such as radiographic overlapping, marginal restoration defects, and cervical artifacts, were the main sources of misclassification. Conclusions: AI algorithms, regardless of architecture or commercial status, showed similar diagnostic capabilities and a conservative detection profile, favoring specificity over sensitivity. Improvements in dataset diversity, labeling precision, and explainability may further enhance reliability for secondary caries detection. Clinical Significance: AI-based CDSSs assist clinicians by providing consistent detection. Their high specificity is particularly valuable in minimizing unnecessary invasive treatments (overtreatment), though they should be used as adjuncts rather than a replacement for expert judgment.

Patterns in the vegetation across arid and semiarid regions may be explained as a form of self-organization driven by water scarcity, and are often modeled through reaction-diffusion dynamics. Recent work has shown that similar mathematical models generate patterns on networks. However, these studies have focused on idealized topologies with no reference to natural pattern-forming systems. Our study aims at bridging these two fields: we employ a physical reaction-diffusion vegetation model, and gradually modify the topology of the diffusion network by adding random shortcuts over a 2-dimensional grid, interpolating between a regular lattice and a random network. We found that network topology strongly shapes both the resulting vegetation patterns and the precipitation range that supports them. Three behavioral regimes emerge. On a regular lattice, high-regularity patterns develop reflecting local diffusion processes. On a random network, the system is dominated by global pressure towards homogenization yielding either a uniform state or a single patch. In the intermediate shortcut density range, as the network topology resembles a small world network, the interaction between the two scales of diffusion generates two kinds of disordered patterns: low-regularity patterns with a well-defined characteristic wavelength, and irregular patterns characterized by a broad patch size distribution. These disordered patterns resemble real-world observations and, in our model, they show different responses to changing precipitation. Although we focused on dryland vegetation, we suggest that network-mediated diffusion could lead to similar mechanisms in a wide variety of pattern-forming systems. HighlightsO_LIWe study vegetation pattern formation over different diffusion network topologies. C_LIO_LITwo kinds of stable disordered patterns states develop over small world topologies. C_LIO_LILow-regularity patterns with a well-defined characteristic wavelength. C_LIO_LIIrregular patterns characterized by a broad patch size distribution. C_LIO_LIThese different kinds of disordered states show different relations to precipitation. C_LI

Knowing the mortality rates associated with infection by a pathogen is essential for effective preparedness and response. Here, harnessing the flexibility of a Bayesian approach, we produce an estimate of the Infection Fatality Ratio (IFR) for A(H5N1) conditional on explicit assumptions, and quantify the uncertainty thereof. We also apply the method to first-wave COVID-19 data up to March 2020, demonstrating the estimates that could be obtained were the model available then. Our analysis uses World Development Indicators (WDI) from the World Bank, the A(H5N1) WHO confirmed cases and deaths tracker by country (2003-2024), and COVID-19 cases and deaths data from John Hopkins University (January and February 2020). Since infectious disease dynamics are typically influenced by local socio-economic factors rather than political borders, individual countries are placed within clusters of countries sharing similar WDIs relevant to respiratory viral diseases, with clusters derived by performing Hierarchical Clustering. To estimate the IFR, we fit a Negative Binomial Bayesian Hierarchical Model for A(H5N1) and COVID-19 separately. We explicitly modelled key unobserved parameters with informative priors from expert opinion and literature. By modelling underreporting, our analysis suggests lower fatality (15.3%) compared to WHO's Case Fatality Ratio estimate (54%) on lab-confirmed cases. However, credible intervals are wide ([0.5%, 64.2%] 95% CrI). Therefore, good preparedness for a potential A(H5N1) pandemic implies adopting scenario planning under our central estimate, as well as for IFRs as high as 70%. Our approach also returns a COVID-19 IFR estimate of 2.8% with [2.5%, 3.1%] 95% CrI which is consistent with literature.

Tuberculosis (TB) is prevalent in Uganda and overlaps with a high rate of HIV/TB coinfection. While nearly all hospital-based TB cases in Kampala, the capital of Uganda, show clear TB symptoms, 30% or more of undiagnosed TB cases found through active screening are asymptomatic. Additionally, the host risk factors for TB in Kampala cannot be distinguished from environmental risk factors. These TB-specific challenges are just part of the complexity, especially in areas with high HIV/AIDS burden. Data science techniques, especially Artificial Intelligence (AI) and Machine Learning (ML) algorithms, could help untangle this complexity by identifying factors related to the host, pathogen, and environment, which are difficult to explain or predict with traditional/conventional methods. In this project, we will use health data science approaches (AI/ML) to identify factors driving TB transmission within households and reasons for anti-TB treatment failure. We will utilize the computational resources at Makerere University and available demographic, clinical, and laboratory data from TB patients and their contacts to develop AI and ML algorithms. These will aim to: (1) identify patients at baseline (month 0) unlikely to convert their sputum or culture results by months 2 and 5, thus at risk of failing TB treatment; (2) identify household contacts of TB cases who are at risk of developing TB disease, as well as contacts who may resist TB infection despite repeated exposure to M. tuberculosis. Achieving these objectives will provide evidence that data science methods are effective for early detection of potential TB cases and high-risk patients, thereby helping to reduce TB transmission in the community. The study protocol received approval from the School of Biomedical Sciences IRB, protocol number SBS-2023-495.

BackgroundAlpha satellites, a superfamily of AT-rich tandem repeats, are the primary DNA component of centromeres in Platyrrhini and Catarrhini. Analyses of the human genome suggest that centromeres behave like biological ridges, with new alpha satellite families expanding at the centromere core, splitting and displacing older ones towards the pericentromeres. The Cercopithecini tribe, which displays an unusual chromosomal evolution involving multiple chromosomal fissions and centromere formations, represents a promising model to enhance our understanding of alpha satellite DNA evolutionary history. We previously applied targeted sequencing to centromere DNA from two distant species drawn from the Cercopithecini terrestrial and arboreal lineages, and characterized six alpha satellite families exhibiting varying mean sequence identities. MethodsCombining classical and molecular cytogenetics, we mapped the chromosomal distribution of these alpha satellite families across 13 Cercopithecini, one Papionini, and one Colobinae species. A nuclear marker-based phylogeny provided an evolutionary framework for interpretation. ResultsOur phylogeny identifies the terrestrial and arboreal lineages, and a newly designated swamp clade. We observed significant interspecies variations in alpha satellite patterns, including differences in presence/absence and distinct chromosomal distribution patterns (centromeric, pericentromeric, or subtelomeric). Families previously described as heterogeneous (83-87% mean sequence identity) exhibit a centromeric position in the swamp lineage, which is characterized by conserved karyotypes. In contrast, these families show a pericentromeric distribution in the terrestrial and arboreal lineages, replaced at the centromere core by more homogeneous families (95-98% mean sequence identity). In the arboreal clade, which is characterized by highly fissioned karyotypes, putative evolutionary new centromeres show a unique co-occurrence of highly homogeneous and heterogeneous families. Conclusion & ImplicationsWe propose a comprehensive evolutionary scenario for alpha satellite DNA in Cercopithecini, where younger families arise at the centromere core, shift toward the pericentromeres as they age, and eventually face extinction. Our study suggests that alpha satellite DNA and chromosomes evolve in an interdependent manner, with satellite diversification and displacement occurring in parallel with chromosome fissions and centromere repositioning. This comparative cytogenomic approach provides both support for the human-based evolutionary model for alpha satellite DNA and novel temporal insights into its diversification dynamics. Beyond evolutionary genomics, our findings highlight the potential of alpha satellite DNA to complement systematic studies in deciphering complex primate evolutionary histories.

Drowning remains a major global public health challenge, yet how built environment characteristics shape population-level drowning risk remains poorly understood. This study linked satellite-derived built environment data to subnational drowning mortality estimates across 203 regions in 12 countries from 2006-2021. It found that built environment associations with drowning mortality are complex, non-linear, and shaped by development context. Urban extent was strongly protective, while built area near water showed protection overall but increased risk when combined with high population crowding. Almost all drowning mortality variance occurred between regions rather than within regions over time, indicating risk is predominantly determined by place-based characteristics. Income-stratified analyses revealed profound heterogeneity: crowding was protective in low-to middle-income settings but near-null in high-income regions, while waterfront development captured very different realities across contexts. These findings highlight the importance of tailoring drowning prevention strategies to local built environment configurations and development contexts.

Increasing human activities along the US west coast are of concern for populations of cetaceans and particularly for a number of large whale species that are recovering from overexploitation during the era of commercial whaling. New rapid monitoring tools, such as satellite imagery analysis powered by recent advances in artificial intelligence, have potential to provide additional broad-scale and near real-time capacities for survey and monitoring. This paper investigates and demonstrates the feasibility of automatic detection of gray whales in sub-meter satellite imagery off the coast of California, USA. Observations and statistical analysis of regional imagery allowed not only an assessment of their detectability but also the development of robust signal processing and machine learning-based solutions for automated detection. To that end, a regional dataset of 221 gray whales was created using signal processing to inform a deep-learning-based detection framework, and 20 different large neural network architectures for feature extraction followed by a support vector machine algorithm for classification were evaluated for their detection performance. Neural network backbones included 19 convolutional neural networks and 1 transformer network. The best architecture generally achieved satisfying performance with an average balanced accuracy reaching up to 99.90%. It is also demonstrated that panchromatic imagery, in spite of the lesser amount of information provided, can be used to perform detection with a relatively high accuracy of 87.05%, allowing wider spatial and temporal coverage. Large-scale deployment of the best performing models over a broad range of regional satellite imagery resulted in the detection of 3353 gray whales, as well as opportunistic detections of humpback, blue and fin whales, in and going from December 28th 2009 to March 26th 2023. It also provided meaningful data points concerning the migration routes of gray whales within the Channel Islands and Southern California Bight. The large number of high-confidence detections indicates the capacity for a large-scale monitoring approach to support state and federal conservation policies such as gear mitigation, vessel speed reduction programs, or shipping lane redefinition that could also be expanded to other areas and for other species.

Canadas ambition for mineral security and its responsibilities to protect at-risk species and uphold Indigenous rights clash in the case of the Grays Bay Road and Port (GBRP) in Nunavut, an infrastructure project intended to unlock critical mineral deposits. We compiled Indigenous and Western science through a density analysis of caribou harvesting data near the proposed project site. We identified three consistently used harvesting hotspots, with the most significant hotspot lying directly in the path of the proposed GBRP project. These results indicate that the GBRP project will have significant and unmitigable negative effects on caribou conservation, food security, and Inuit harvesting rights. Prime Minister Carney claims that middle power countries must act consistently in this era of geopolitical rupture; this commitment must transfer to natural resource development reviews so that decision-making may be consistent and rooted in cross-legislation responsibilities and values, including the land claims agreements between Indigenous groups and the Government of Canada.

Continuous, non-invasive cardiovascular monitoring is limited by the superficial sensing depth of Photoplethysmography (PPG), which is susceptible to peripheral artifacts. This study evaluates a wearable dual-modality prototype integrating dryelectrode Impedance Plethysmography (IPG) and PPG within a smartwatch form factor. Results from a pilot study (N=2) demonstrate that IPG signals exhibit a temporal lead over PPG across ventral and dorsal sites, supporting its greater penetration depth. During brachial artery modulation, IPG showed superior sensitivity to arterial recovery on the ventral forearm. Furthermore, 60-minute napping sessions revealed that while PPG remained morphologically stable, IPG signals underwent significant evolution, capturing distinct pulsewave archetypes. These findings suggest that wearable IPG provides a high-fidelity window into deep systemic hemodynamics typically reserved for clinical instrumentation.

BackgroundIn patients requiring respiratory support, clinicians rely on physical exam, radiologic, laboratory, and ventilator-derived measures for the provision of sufficient support while minimizing ventilator and "work of breathing" induced lung injury. Point of care lung ultrasound (LUS) is a widely available tool in hospital and clinic environments. To date, LUS has not been used to evaluate lung strain. MethodsWe collected LUS images in four anesthetized, neuromuscularly blocked, and mechanically ventilated pigs being used for another experiment. A feature tracking tool was developed which tracked echo-bright lung structures in ten second clips obtained in triplicate of the right and left, upper and lower lung fields using tidal volumes of 4, 6, 8, 10, and 12 mL/kg. Pleural lines were manually drawn and a program for quantifying lung strain developed with assistance from Anthropic Claude Artificial Intelligence tool. Structures were identified in inspiratory and expiratory frames and tracked bidirectionally with median strain per frame used for calculations. ResultsTriplicate measures of lung ultrasound images in four pigs had a median coefficients of variation of 35% (23-47% IQR) and linear modeling of strain with tidal volumes of 4-12 mL/kg showed positive correlation with R2 value ranging from 0.89 to 0.97. Strain measurements were similar after bronchial administration of 1.5M hydrochloric acid. ConclusionsRegional lung strain quantification using LUS is a viable and potentially useful tool for respiratory support management.

Diet is an important ecological modulator of the oral microbiome, yet population-level evidence on a broader spectrum of food components remains limited. This cross-sectional study investigated associations among dietary intake, oral rinse microbiome, and oral disease conditions in a nationally representative sample of United States adults from the National Health and Nutrition Examination Survey. A total of 3,254 participants with oral rinse microbiome sequencing data were included, with oral conditions classified as oral health, caries-only, periodontitis-only, or co-existing disease. Dietary intake was assessed using 24-hour dietary recalls and summarized as dietary indices and energy-adjusted food components. Associations between diet and the oral microbiome were evaluated using community-level analyses, regression models, mediation analyses, and unsupervised clustering, while accounting for oral conditions. This study found that dietary intake, as a combined variable set, explained 3.6% of the variance in oral rinse microbial community structure; this was comparable to oral disease status or smoking and larger than sociodemographic factors. Healthier dietary profiles, including higher health-associated dietary index scores and greater vegetable and fruit intake, were associated with taxa commonly linked to oral health (e.g., Neisseria, Cardiobacterium and Lautropia). In contrast, added sugars, alcoholic drinks, cured meat, potatoes, dairy products, and higher dietary inflammatory index scores showed opposite association patterns. Mediation analyses suggested that coordinated microbial groups may partly link dietary exposures with oral disease outcomes, particularly for vegetables and added sugars. Additionally, three population-level dietary patterns were identified, among which the plant-rich pattern was associated with more favorable oral health and microbial profiles enriched in nitrate-reducing commensals, including Neisseria and Haemophilus. Overall, dietary intake was associated with oral microbiota composition and oral health conditions, supporting ecological influences of dietary components beyond sugar on oral bacteria and dental diseases. Longitudinal studies are needed to clarify the direction and causality of these relationships.

Tick-borne diseases are now reported from nearly every county in Illinois, and three vector tick species (Amblyomma americanum, Dermacentor variabilis, and Ixodes scapularis) are of particular concern because these are responsible for most of the tick-borne disease transmission in the state. However, active surveillance is patchy, many counties have little or no sampling, and there is no statewide, quantitative map of relative abundance that can be used to anticipate risk in unsampled areas. To address these gaps, we developed Bayesian hierarchical spatial models to estimate the county-level abundance of these three vector tick species in Illinois. Using active surveillance data from 2019-2022, we modeled county-level abundance as a function of climate, land cover, forest fragmentation, and deer habitat suitability. Spatial dependence was captured using a Besag-York-Mollie 2 (BYM2) prior implemented in INLA, along with spatial 5-fold cross-validation to assess predictive performance. A. americanum showed the highest predicted abundance in southern and central Illinois, D. variabilis was widespread but diffuse, and I. scapularis was concentrated in northern and selected central counties. Together, these models provide the first spatial, statewide, uncertainty-aware assessment of tick abundance in Illinois, highlighting priority counties where surveillance lags disease risk.

Enhanced resilience conferred through sublethal stress pre-exposure may be crucial for reef building corals to cope with variable environments. The effect of stress priming on Acropora cervicornis thermotolerance was evaluated in the context of elevated temperature and ammonium enrichment, 3 and/or 6 M above ambient, respectively. Primed corals were pre-exposed to each stressor individually or in combination for eight days, while non-primed corals remained at ambient conditions. After an eight-day recovery, primed corals and a subset of non-primed corals (naive) were subjected to an acute 15-hour thermal challenge. Coral metabolism, symbiosis, and gene expression were characterized throughout the experiment. Thermal tolerance was quantified as algal symbiont, chlorophyll, and live tissue retention, along with survival probability following acute heating. Primed corals were more likely to retain symbionts and chlorophyll after heat stress and also exhibited slower tissue loss. Moreover, thermal pre-exposure reduced the risk of tissue loss or predicted mortality. Apoptotic regulation differed between primed and naive corals during the initial and secondary heat exposures. Additionally, primed corals exhibited patterns of transcriptional resilience under acute thermal stress. Altogether, results provide support for the capacity of A. cervicornis to gain resilience through pre-exposure to ecologically relevant conditions as well as insights into the molecular mechanisms underpinning this process.

There is more to musculotendon path modeling than aligning a cable to reflect the geometric features of a muscle-tendon unit. From the perspective of simulation accuracy, the key is to replicate the length- and moment arm-joint angle relations of the target muscle. In this study, we propose an effect-oriented approach of automated path modeling, via the hybrid calibration based on muscle surface mesh and moment arm. The task is formulated as an optimization problem with a threefold objective for the path to: 1) pass through multiple ellipses representing muscle cross-sections, 2) yield moment arms that match experimental measurements, and 3) yield moment arms with the designated signs. The performance of our optimization framework is demonstrated with the musculoskeletal surface mesh from the Visible Human Male and moment arm datasets from literature--producing 42 paths that are anatomically realistic and biomechanically accurate in 20.1 min. Our optimization framework is gradient-specified, which is faster and more accurate than using the default numerical gradient, making it applicable for large-scale subject-specific uses.

This paper investigates the way in which mitochondria distribute and align inside HeLa cells observed with serial block-face scanning electron microscopy. Four models of alignment were considered: (1) mitochondria exhibiting no discernible alignment pattern, (2) mitochondria aligned pointing towards the nucleus of the cell, (3) mitochondria aligned all in one direction when viewed from above, (4) mitochondria aligned tangent to the surface of the nucleus. These models were named (1) unaligned, (2) petals, (3) racecars, and (4) clouds. The mitochondria, nucleus and plasma membrane of 25 individual cells were segmented. A total of 12,299 mitochondria were identified and analysed. Alignment of the major axis of each mitochondrion was calculated in two ways: relative to a ray that joins it to the centroid of the nucleus, and relative to a ray that joins it to the nucleus surface. Results indicate that mitochondria tend to align tangentially to the nucleus surface, i.e., a clouds model. In addition, differences in the spatial distributions of the mitochondria were found and quantified with clearly defined metrics. The methodology here presented can be extended to other acquisition settings where the distribution and alignment of cells could be important, for instance, histopathology.

Mass mortality of reef-building stony corals has driven widespread community shifts towards reefs dominated by soft corals and macroalgae. Although physical competition for space between these organisms plays an important role, non-contact water-mediated interactions have been proposed to modulate organismal performance and community functioning, yet their independent effects remain poorly resolved. Here, we experimentally tested the hypothesis that water-mediated interactions generate non-additive effects on community productivity, altering ecosystem functioning during phase shifts. Using two controlled incubation experiments with representative stony corals, soft corals, and macroalgae, we compared monoculture baseline productivity with mixed assemblages across a gradient of biomass ratios mimicking phase shift scenarios. We found that reductions in stony coral biomass led to community-level declines in photosynthesis and calcification that exceeded expectations based on monocultures, indicating emergent negative effects of community restructuring. However, these effects were strongly species-dependent, with some assemblages showing only minor deviations from expectations, whereas others exhibited pronounced productivity losses. At the species level, both stony corals reduced photosynthetic efficiency in mixed assemblages, while soft corals maintained efficiency across treatments. Macroalgal responses diverged, with one species exhibiting reduced and another increased photosynthetic efficiency in mixed communities. These species-specific physiological responses scaled up to explain community-level deviations from expected productivity, suggesting that gains in productivity by certain taxa can partially offset, but not fully compensate for, losses in coral-driven functions such as calcification. Together, our findings indicate that sublethal, water-mediated interactions can reorganize holobiont functioning and lead to changes in ecosystem productivity, independent of direct physical competition. By altering community-wide energy acquisition and carbonate production, such interactions may reinforce feedback loops that accelerate ecosystem phase shifts. We argue that incorporating water-mediated interaction effects into ecological theory and ecosystem models is essential for predicting the stability and recovery potential of coral reefs and other transitioning ecosystems under climate change.