A declared shelf life aligned with consumer acceptance is essential to avoid unnecessary food waste. The relationship between microbial development and consumers' odour-based decisions to accept or discard industrially packed chicken breasts from three European countries (Hungary, Norway and Portugal) was investigated. Products were stored at 4 °C up to 27 days, with or without a shift to 8 °C after 3-7 days to mimic temperature abuse at retail/consumer stages. Bacterial levels were quantified by culture-based enumeration (total viable counts, TVC), and the microbiota assessed by 16S rRNA gene amplicon sequencing. In each country ≥120 consumers assessed acceptance based on odour. Acceptance was modelled through survival analysis using time (days) and TVC as predictors. Initial TVC differed among products (2.5 - 4.3 log cfu/g). Dominant taxa were Photobacterium (Hungary), Lactobacillales (Norway), and Brochothrix together with Yersiniaceae (Portugal) at the end of the declared shelf life. Temperature abuse shifted the microbiota towards higher proportions of Enterobacterales (Norway and Portugal). Sensory shelf life varied markedly, with 25% rejection rates reached after 9 days (Hungary), 25 days (Norway), and 12 days (Portugal) during storage at 4 °C, corresponding to TVCs of 6.4, 7.4 and 7.3 log cfu/g. We propose a consumer-anchored methodological approach for shelf life assessment that (i) predefines an acceptable rejection level (ii) models acceptance as a function of TVC under foreseeable abuse and (iii) complements culture-dependent analyses with culture-independent profiling. This approach links microbial criteria to consumer perception and behaviour and providing a basis for more accurate, consumer-anchored shelf-life determination.

Refining approaches to measuring, monitoring and appraising animal welfare in aquaculture research is key to (i) protecting and optimizing it, (ii) documenting the severity of how and when it deviates, and (iii) ensuring good scientific quality, reliable results and reproducibility, amongst other factors. However, different fish species and life stages can have varying welfare needs and assessing their welfare can be challenging. An array of welfare indicators (WIs) can be utilized when documenting fish welfare, and there is currently little consensus on which WIs are most applicable to the key fish species used in European aquaculture research. The aim of this review is to propose updated, fit for purpose and comprehensive WI toolboxes for aquaculture research involving Atlantic salmon ( Salmo salar ), rainbow trout ( Oncorhynchus mykiss ), European seabass ( Dicentrarchus labrax ), gilthead seabream ( Sparus aurata ), and the common carp ( Cyprinus carpio ). Where possible, these toolboxes will also include life-stage considerations. It also provides information on utilizing WIs in deciding humane end-points as well as information on how to sample different types of indicators. The review closes with information on how digitalization can affect the collection, collation and analysis of WI data in aquaculture research, including both practical and theoretical considerations. The toolboxes incorporate a range of WIs that go beyond those required for legally safeguarding fish welfare in both laboratory and operational experimental facilities in the current European 2010/63/EU Directive on the protection of animals used for scientific purposes and its amendment, the Commission Delegated Directive (EU) 2024/1262.

The CCI package provides a computational framework for conditional independence testing in R, combining machine learning models with Monte Carlo cross-validation to deliver robust error control across complex data structures. CCI is model-agnostic, user-friendly, and supports continuous, categorical, and mixed data. Key func tionalities include automated hyperparameter tuning, flexible direction selection, and visualization tools for null distributions and p-values. By lowering the barrier to rigorous conditional independence testing, the package facilitates advances in causal inference research and applied data analysis.

Despite extensive research on poultry spoilage, variability across batches, seasons, and facilities remains poorly resolved. This limits the ability to determine whether microbiological data can be used to predict sensory shelf life at the batch level, and thereby support adaptive, data-driven shelf-life determination. This study examined how initial microbial status, growth dynamics, microbiota composition, and sensory changes varied in modified atmosphere packed (MAP) chicken breast fillets from two European producers (A, Norway; B, Portugal) sampled over one year. Fillets were analysed after production and during storage at 4 °C or under a 4 to 8 °C temperature shift. Initial total viable counts (TVC) were lower in fillets from Producer A than B, but both producers showed approximately 2-log variation between batches. Maximum growth rates were similar, but fillets from Producer B showed consistently shorter lag times, reaching 7 log10 CFU/g earlier. Lag time was the primary determinant of shelf-life variation, whereas initial aerobic TVC had limited predictive value. During late storage, microbiota communities converged toward producer-specific spoiler profiles that were stable across batches (A: Lactobacillales, Hafnia_Obesumbacterium; B: Brochothrix, Yersiniaceae). Sensory deterioration correlated with TVC and was accelerated but not qualitatively altered by elevated temperature. Overall, initial aerobic TVC was not a reliable predictor of batch-specific shelf life; instead, early measurements that enable estimation of batch-specific lag time (i.e., the fraction of bacteria able to grow in the product) may provide an actionable basis for adaptive shelf-life setting within producers.

Measuring and monitoring fish welfare in aquaculture research relies on the use of outcome- (biotic) and input-based (e.g., abiotic) welfare indicators (WIs). Incorporating behavioural auditing into this toolbox can sometimes be challenging because sourcing quantitative data is often labour intensive and it can be a time-consuming process. Digitalization of this process via the use of computer vision and artificial intelligence can help automate and streamline the procedure, help gather continuous quantitative data and help process optimisation and assist in decision-making. The tool introduced in this study (1) adapts the DeepLabCut framework, based on computer vision and machine learning, to obtain pose estimation of Atlantic salmon parr under replicated experimental conditions, (2) quantifies the spatial distribution of the fish through a toolbox of metrics inspired by the ecological concepts home range and core area, and (3) applies it to inspect behavioural variability in and around feeding. This proof of concept study demonstrates the potential of our methodology for automating the analysis of fish behaviour in relation to home range and core area, including fish detection, spatial distribution and the variations within and between tanks. The impact of feeding on these patterns is also briefly outlined, using 5 days of experimental data as a demonstrative case study. This approach can provide stakeholders with valuable information on how the fish use their rearing environment in small-scale experimental settings and can be used for the further development of technologies for measuring and monitoring the behaviour of fish in research settings in future studies.

The evolving landscape of agri-food systems, driven by climate change and population growth, necessitates innovative approaches to ensure food integrity, safety, and sustainability. This review explores the role of data fusion strategies, particularly focusing on non-destructive spectroscopic sensors (NDSS) in three key application contexts: in-field monitoring, on/in-line food processing, and food quality authentication. Various data fusion scenarios, including fusing spectra from different spectroscopic platforms, integrating images and spectra, and combining non-spectroscopic sensor data with spectroscopic ones are reviewed. Focus is set on practical considerations, such as selecting the level of data fusion, defining blocks, variable selection, and validation methods, highlighting the importance of tailored approaches based on research aims and data characteristics.While combining information from diverse sensors generally enhances information extraction and modeling performance, their implementation in routine applications is still limited and especially studies focused on data fusion models’ performance over time and their maintenance are lacking.

Conditional Independence (CI) testing is fundamental in statistical analysis. For example, CI testing helps validate causal graphs or longitudinal data analysis with repeated measures in causal inference. CI testing is difficult, especially when testing involves categorical variables conditioned on a mixture of continuous and categorical variables. Current parametric and non-parametric testing methods designed for continuous variables and can quickly fall short in the categorical case. This paper presents a computational approach for CI testing for categorical data types, which we call computational conditional independence (CCI) testing. The test procedure is based on permutation and combines machine learning prediction algorithms and Monte Carlo Cross-Validation. We evaluated the approach through simulation studies and assessed the performance against alternative methods; the generalized covariance measure (GCM) test, the kernel conditional independence (KCI) test, and testing with multinomial regression. We find that the computational approach to testing has utility over the alternative methods, and we can achieve better control over type-I-error rates. We hope this work can expand the toolkit for CI testing for practitioners and researchers.

The main objective of this study was to design, build, and test a compact, multi-well, portable dry film FTIR system for industrial food and bioprocess applications. The system features dry film sampling on a circular rotating disc comprising 31 wells, a design that was chosen to simplify potential automation and robotic sample handling at a later stage. Calibration models for average molecular weight (AMW, 200 samples) and collagen content (68 samples) were developed from the measurements of industrially produced protein hydrolysate samples in a controlled laboratory environment. Similarly, calibration models for the prediction of lactate content in samples from cultivation media (59 samples) were also developed. The portable dry film FTIR system showed reliable model characteristics which were benchmarked with a benchtop FTIR system. Subsequently, the portable dry film FTIR system was deployed in a bioprocessing plant, and protein hydrolysate samples were measured at-line in an industrial environment. This industrial testing involved building a calibration model for predicting AMW using 60 protein hydrolysate samples measured at-line using the portable dry film FTIR system and subsequent model validation using a test set of 26 samples. The industrial calibration in terms of coefficient of determination (R2 = 0.94), root mean square of cross-validation (RMSECV = 194 g mol-1), and root mean square of prediction (RMSEP = 162 g mol-1) demonstrated low prediction errors as compared to benchtop FTIR measurements, with no statistical difference between the calibration models of the two FTIR systems. This is to the authors' knowledge the first study for developing and employing a portable dry film FTIR system in the enzymatic protein hydrolysis industry for successful at-line measurements of protein hydrolysate samples. The study therefore suggests that the portable dry film FTIR instrument has huge potential for in/at-line applications in the food and bioprocessing industries.

Industries are continuously looking for dedicated sensor solutions for evaluating the chemical composition of products that can provide valuable insights into process control, process optimization, and product quality. A rapid and robust analytical method, Fourier-Transform Infrared (FTIR) spectroscopy, holds significant potential in this regard as it provides detailed compositional values of food nutrients. The main aim of the present study was to use dry film FTIR spectroscopy as an analytical tool to characterize products from an industrial enzymatic protein hydrolysis process and link this to understand industrial process variations. For this purpose, 463 protein hydrolysate samples were obtained from an industrial enzymatic protein hydrolysis plant. In the same period, systematic variations in process parameters such as raw material composition, enzyme type, and water addition, were performed. All samples were analyzed using dry film FTIR spectroscopy. Two subsets containing 200 and 68 hydrolysate samples were chosen for measuring average molecular weight (AMW) and collagen content respectively, providing reference data for constructing calibration models based on partial least squares regression (PLSR). The percentage of low molecular weight constituents was derived from the molecular weight distribution of size exclusion chromatograms of protein hydrolysates and also used in the modeling. Calibration models for the prediction of AMW, low molecular weight constituents, and collagen content were obtained with a good fit and lower estimation errors. Principal component analysis (PCA) of protein hydrolysates’ FTIR spectra effectively differentiated process variations (enzyme types and raw materials) without extensive reference analysis. One-factor analysis of variance (ANOVA) tests was used to observe the impact of process variation on product quality. FTIR proved to be a sensitive method for liquid protein analysis and process control with a significant potential in process optimization approaches. To the authors’ knowledge, this is the first time dry film FTIR spectroscopy has been used to evaluate an industrial bioprocess with designed process variations.

Conditional Independence (CI) testing is fundamental in statistical analysis. For example, CI testing helps validate causal graphs or longitudinal data analysis with repeated measures in causal inference. CI testing is difficult, especially when testing involves categorical variables conditioned on a mixture of continuous and categorical variables. Current parametric and non-parametric testing methods are designed for continuous variables and can quickly fall short in the categorical case. This paper presents a computational approach for CI testing suited for categorical data types, which we call computational conditional independence (CCI) testing. The test procedure is based on permutation and combines machine learning prediction algorithms and Monte Carlo cross-validation. We evaluated the approach through simulation studies and assessed the performance against alternative methods: the generalized covariance measure test, the kernel conditional independence test, and testing with multinomial regression. We find that the computational approach to testing has utility over the alternative methods, achieving better control over type I error rates. We hope this work can expand the toolkit for CI testing for practitioners and researchers.

When vibrational spectroscopy is used for quantification purposes, multivariate analysis is often used to extract information from covariances between the spectra and any given reference values. In complex samples, there is a high risk that the constituents covary with each other. In such scenarios many methods may confuse the analytes and use signal from several analytes, rather than just the analyte of interest. While this allows the method to use more signal, and thus have a better effective signal-to-noise ratio, it also makes them less robust to changes to the chemical composition in the samples. This effect has been termed the cage of covariance. In order to avoid cage of covariance to affect predictive performances, it is highly important to have simple diagnostic tools to analyze and review this effect. Therefore, in the present paper, a systematic overview of tools for diagnosing and quantifying the cage of covariance in spectroscopic calibration models is provided. A collection of previously published methods with some expansions is provided, as well as two completely new tools: covariance ratio and virtual spiking. Practical applications of the tools on three different datasets are also shown.

Auditing fish welfare within the aquaculture research community relies on the thorough monitoring of outcome- (biotic) and input-based (e.g., abiotic) elements. The community utilises operational and laboratory-based welfare indicators (i.e., OWIs and LABWIs) as fit-for-purpose tools to help with that task. Incorporating behavioural auditing in this toolbox can sometimes be challenging because sourcing quantitative data is often labour intensive, and it can be a time-consuming process. Digitalization of this auditing via the use of computer vision and artificial intelligence can help automate the monitoring process, help gather continuous quantitative data and help process optimisation and the decision-making process. This study i) adapts the DeepLabCut framework, based on computer vision and machine learning, to obtain pose estimation of Atlantic salmon parr under replicated experimental conditions, ii) quantifies the fish spatial distribution through a toolbox of metrics inspired by the ecological concepts home range and core area, and iii) applies it to inspect behavioural variability in and around feeding. These results demonstrate that the aforementioned methodology can automate the auditing of i) fish detection, ii) spatial distribution of fish in tanks and iii) the variability of spatial distribution before, during and after feeding within and between tanks and days. This can provide stakeholders with valuable information on how the fish use their rearing environment and can be used to inspire further development of technologies for behavioural auditing of fish in research settings.

This study introduces a novel computational approach for testing conditional independence (BB CI test) within causal Directed Acyclic Graphs (DAGs), leveraging Bayesian non-parametric bootstrap and machine learning techniques. Our method offers an alternative for validating the assumptions underpinning causal DAGs. Through simulation studies and an industrial case analysis, we demonstrate the test procedure in accurately assessing conditional independence, comparing it with the Generalized Covariance Measure (GCM) test. Our findings suggest that the BB CI test is advantageous in scenarios where existing methods may falter due to violations of model assumptions. This research contributes to the causal inference literature by providing a computational tool for researchers and practitioners to validate causal models.

The study introduces three novel strategies for incorporating capabilities for dynamic modelling into multiblock regression methods by integrating sequentially orthogonalised partial least squares (SO-PLS) with different dynamic modelling techniques. The study evaluates these strategies using synthetic datasets and an industrial example, comparing their performance in predictive ability, identification of process dynamics, and quantification of block contributions. Results suggest that these approaches can effectively model the dynamics with performance comparable to state-of-the-art methods, providing, at the same time, insight into the dynamic order and block contributions. One of the strategies, sequentially orthogonalised dynamic augmented (SODA)-PLS, shows promise by ensuring that redundant information in the time dimension is not included, resulting in simpler and more easily interpretable dynamic models. These multiblock dynamic regression strategies have potential applications for improved process understanding in industrial settings, especially where multiple data sources and inherent time dynamics are present.

IntroductionAvian eggshell membrane (ESM) is a complex extracellular matrix comprising collagens, glycoproteins, proteoglycans, and hyaluronic acid. We have previously demonstrated that ESM possesses anti-inflammatory properties in vitro and regulates wound healing processes in vivo. The present study aimed to investigate if oral intake of micronized ESM could attenuate skeletal muscle aging associated with beneficial alterations in gut microbiota profile and reduced inflammation.MethodsElderly male C57BL/6 mice were fed an AIN93G diet supplemented with 0, 0.1, 1, or 8% ESM. Young mice were used as reference. The digestibility of ESM was investigated using the static in vitro digestion model INFOGEST for older people and adults, and the gut microbiota profile was analyzed in mice. In addition, we performed a small-scale pre-clinical human study with healthy home-dwelling elderly (>70 years) who received capsules with a placebo or 500 mg ESM every day for 4 weeks and studied the effect on circulating inflammatory markers.Results and discussionIntake of ESM in elderly mice impacted and attenuated several well-known hallmarks of aging, such as a reduction in the number of skeletal muscle fibers, the appearance of centronucleated fibers, a decrease in type IIa/IIx fiber type proportion, reduced gene expression of satellite cell markers Sdc3 and Pax7 and increased gene expression of the muscle atrophy marker Fbxo32. Similarly, a transition toward the phenotypic characteristics of young mice was observed for several proteins involved in cellular processes and metabolism. The digestibility of ESM was poor, especially for the elderly condition. Furthermore, our experiments showed that mice fed with 8% ESM had increased gut microbiota diversity and altered microbiota composition compared with the other groups. ESM in the diet also lowered the expression of the inflammation marker TNFA in mice and in vitro in THP-1 macrophages. In the human study, intake of ESM capsules significantly reduced the inflammatory marker CRP. Altogether, our results suggest that ESM, a natural extracellular biomaterial, may be attractive as a nutraceutical candidate with a possible effect on skeletal muscle aging possibly through its immunomodulating effect or gut microbiota.

Digital sensors and machine learning enable efficiency improvements in production processes, through process monitoring, anomaly detection, soft sensing, and process control. However, the development of such solutions requires several data preprocessing steps. In continuous processes, a crucial part of the data preparation is adjusting for time delays between different sensors. This is necessary to ensure that each sensor measurement relate to the same volume of materials going through various processing steps.This study provides an overview of data-driven methods for estimating time lags between sensors in continuous processes. The methods are assessed in a large simulation study, on data sets with different sample sizes, model complexities and autocorrelation functions. Our results shows that most methods work well if the relationships are close to linear, but more flexible metrics like distance correlation and maximum information coefficient are needed in more complex systems. Finally, we present a real industrial example to illustrate some real-world aspects of the variable time delay estimation process.

There is an increasing interest in detecting and describing causal pathways that link different features of a system together, usually referred to as path modelling or path analysis. The features are often represented by multiple descriptor variables, requiring multivariate data fusion methods that take the pathways between groups of variables into account. Interestingly, this takes us back to the origins of Partial Least Squares by Herman Wold, as a “soft” alternative to covariance-based structural equation modelling. During the 16th Scandinavian Symposium on Chemometrics (SSC16) in 2019, there was a session dedicated to “Path modeling, graphical modeling and causality” in which several methodological improvements to PLS path modelling were discussed: PLS path modelling has some serious drawbacks when it comes to modelling data from complex systems, most remarkably that the effects describing the pathways need to remain very simple. During the intermittent corona period, several researchers developed improvements to these limitations. The meeting was organized by Jeroen Jansen, Age Smilde and Ingrid Måge and was held in the beautiful city of Zaandam, close to Amsterdam. All 15 participants were specially invited due to their interest and expertise in different aspects of path modelling of multivariate data. The participants represented seven different institutions in Italy, the Netherlands and Norway and quantitative methodology in several research disciplines: life sciences, food research, chemical engineering and social science. The aim of the meeting was to discuss methods, challenges and applications, to learn from each other and to foster collaborations. The format of the meeting was informal, and the participants openly shared new ideas and engaged in energetic and fruitful discussions. The first day started with extensive overview presentations, followed by presentations of recent advances, including some discussion, by all participants. The second day was spent purely on discussions, ranging from in-depth discussions on methodology to a quo vadis on the next generation of required innovations. All participants had prepared a presentation, where they were asked to focus on questions and problems rather than success stories. Age Smilde (University of Amsterdam) started by giving an overview of existing methods for path modelling. He also presented a list of proposed requirements for methods and indicated the status of the existing methods relative to these requirements. Rosaria Romano (University Federico II of Naples) gave an overview of PLS path modelling and various extensions of the method. Johan Westerhuis (University of Amsterdam) presented the method iTop (“inferring the topology of omics data”), a recently proposed method for causal discovery based on partial matrix correlations. Anna Heintz-Buschart and Roel van der Ploeg (both University of Amsterdam) presented challenging applications of path modelling in ecology and biostatistics. Kristian Hovde Liland (Norwegian University of Life Sciences) gave an overview of methods for calculating (partial) matrix correlations. Tormod Næs (Nofima) presented the SO-PLS path modelling method and some new ideas for estimating direct and indirect effects. Ingrid Måge (Nofima) presented the use of Directed Acyclic Graphs (DAG's) in SO-PLS path modelling and in causal modelling. Lars Erik Solberg (Nofima) discussed the relationship between theory and data and to what extent a poor fit should make us discard a causal model. Christian Thorjussen (Nofima/Norwegian University of Life Sciences) presented thoughts about how to check testable implications of a DAG, for different data types (continuous, binary and multi-class). Geert van Kollenburg (Eindhoven University of Technology) and Tim Offermans (Radboud University) presented their process-PLS method for path modelling, including some nice applications, challenges and plans for future developments. Sin Yong Teng (Radboud University) presented how all current methods suffer from a restriction due to DAGs, where feedback loops cannot be explicitly modelled. Carlo Bertinetto (Teijin Aramid) presented how he has a considerable need for path models in the process industry, to understand and optimize chemical processes. All participants agreed that the meeting was inspiring, led to deeper insight into various aspects of path modelling and strengthened the connection between different research groups. Although there were participants with clearly distinctive backgrounds, the mutual desire for understanding, respect for complementary standpoints and patience in listening and delivering argumentation that is so familiar from chemometrics meetings made the meeting invaluable, informative and constructive. We agreed on two concrete action points. The first one is to collaborate on a paper that compares existing approaches for a range of different data examples, to highlight the usability, strengths and weaknesses of the methods in different scenarios. The second action point is to organize a larger follow-up meeting; the current meeting already gathered a Pan-European community of differential experience and scientific background. The discussions, however, raised already some highly fundamental points for further development that requires expertise that is partly complementary to that gathered in the Zaandam meeting. If there is interest in joining such a larger gathering, the reader is encouraged to contact any of the organizers. The peer review history for this article is available at https://publons.com/publon/10.1002/cem.3430. The peer review history for this article is available at https://publons.com/publon/10.1002/cem.3430.