Probabilistic topic modelling in food spoilage analysis: A case study with Atlantic salmon (Salmo salar).

Probabilistic topic modelling is frequently used in machine learning and statistical analysis for extracting latent information from complex datasets. Despite being closely associated with natural language processing and text mining, these methods possess several properties that make them particularly attractive in metabolomics applications where the applicability of traditional multivariate statistics tends to be limited. The aim of the study was thus to introduce probabilistic topic modelling - more specifically, Latent Dirichlet Allocation (LDA) - in a novel experimental context: volatilome-based (sea) food spoilage characterization. This was realized as a case study, focusing on modelling the spoilage of Atlantic salmon (Salmo salar) at 4 °C under different gaseous atmospheres (% CO2/O2/N2): 0/0/100 (A), air (B), 60/0/40 (C) or 60/40/0 (D). First, an exploratory analysis was performed to optimize the model tunings and to consequently model salmon spoilage under 100% N2 (A). Based on the obtained results, a systematic spoilage characterization protocol was established and used for identifying potential volatile spoilage indicators under all tested storage conditions. In conclusion, LDA could be used for extracting sets of underlying VOC profiles and identifying those signifying salmon spoilage, giving rise to an extensive discussion regarding the key points associated with model tuning and/or spoilage analysis. The identified compounds were well in accordance with a previously established approach based on partial least squares regression analysis (PLS). Overall, the outcomes of the study not only reflect the promising potential of LDA in spoilage characterization, but also provide several new insights into the development of data-driven methods for food quality analysis.

Spoilage evaluation of raw Atlantic salmon (Salmo salar) stored under modified atmospheres by multivariate statistics and augmented ordinal regression.

The development of quality monitoring systems for perishable food products like seafood requires extensive data collection under specified packaging and storage conditions, followed by advanced data analysis and interpretation. Even though the benefits of using volatile organic compounds as food quality indices have been recognized, few studies have focused on real-time quantification of the seafood volatilome and subsequent systematic identification of the most important spoilage indicators. In this study, spoilage of Atlantic salmon (Salmo salar) stored under modified atmospheres (% CO2/O2/N2) and air was characterized by performing multivariate statistical analysis and augmented ordinal regression modelling for data collected by microbiological, chemical and sensory analyses. Out of 25 compounds quantified by selected-ion flow-tube mass spectrometry, ethanol, dimethyl sulfide and hydrogen sulfide were found characteristic under anaerobic conditions (0/0/100 and 60/0/40), whereas spoilage under air was primarily associated with the production of alcohols and ketones. Under high-O2 MAP (60/40/0), only 3-methylbutanal fulfilled the identification criteria. Overall, this manuscript presents a systematic and widely applicable methodology for the identification of most potential seafood spoilage indicators within the context of intelligent packaging technology development. In particular, parallel application of statistics and modelling was found highly beneficial for the performance of the quality characterization process and for the practical applicability of the obtained results in food quality monitoring.

Assessing the potential of the qualitative trajectory calculus to detect gait pathologies: a case study of children with developmental coordination disorder.

The Qualitative Trajectory Calculus (QTC) is a qualitative spatio-temporal calculus for describing interactions between moving point objects. So far, it remained unclear whether QTC is useful for describing subtle differences, such as between the movements of different parts of a human body. We tested the applicability of QTC to detect differences in the gait patterns of children with or without Developmental Coordination Disorder (DCD). We found that using a combination of three markers (i.e. ankle, toe and trochanter), QTC can achieve a high classification accuracy (i.e. 83.3%) of classifying subjects correctly to either the DCD group or the control group.

Microbiological, chemical and sensory spoilage analysis of raw Atlantic cod (Gadus morhua) stored under modified atmospheres.

During fish spoilage, microbial metabolism leads to the production of volatile organic compounds (VOCs), characteristic off-odors and eventual consumer rejection. The aim of the present study was to contribute to the development of intelligent packaging technologies by identifying and quantifying VOCs that indicate spoilage of raw Atlantic cod (Gadus morhua) under atmospheres (%v/v CO2/O2/N2) 60/40/0, 60/5/35 and air. Spoilage was examined by microbiological, chemical and sensory analyses over storage time at 4 or 8 °C. Selected-ion flow-tube mass spectrometry (SIFT-MS) was used for quantifying selected VOCs and amplicon sequencing of the 16S rRNA gene was used for the characterization of the cod microbiota. OTUs classified within the Photobacterium genus increased in relative abundance over time under all storage conditions, suggesting that Photobacterium contributed to spoilage and VOC production. The onset of exponential VOC concentration increase and sensory rejection occurred at high total plate counts (7-7.5 log). Monitoring of early spoilage thus calls for sensitivity for low VOC concentrations.

Predictions of methane emission levels and categories based on milk fatty acid profiles from dairy cows.

Milk fatty acid (MFA) have already been used to model methane (CH4) emissions from dairy cows. However, the data sets used to develop these models covered limited variation in dietary conditions, reducing the robustness of the predictions. In this study, a data set containing 140 observations from nine experiments (41 Holstein cows) was used to develop models predicting CH4 expressed as g/day, g/kg dry matter intake (DMI) and g/kg milk. The data set was divided into a training (n=112) and a test data set (n=28) for model development and validation, respectively. A generalized linear mixed model was fitted to the data using the marginal R 2 (m) and the Akaike information criterion to evaluate the models. The coefficient of determination of validation (R 2 (v)) for different models developed ranged between 0.18 and 0.41. Form the intake-related parameters, only inclusion of total DMI improved the prediction (R 2 (v)=0.58). In addition, in an attempt to further explore the relationships between MFA and CH4 emissions, the data set was split into three categories according to CH4 emissions: LOW (lowest 25% CH4 emissions); HIGH (highest 25% CH4 emissions); and MEDIUM (50% remaining observations). An ANOVA revealed that concentrations of several MFA differed for observations in HIGH compared with observations in LOW. Furthermore, the Gini coefficient was used to describe the MFA distribution for groups of MFA in each CH4 emission category. The relative distribution of the MFA, particularly of the odd- and branched-chain fatty acids and mono-unsaturated fatty acids of observations in category HIGH differed from those in the other categories. Finally, in an attempt to validate the potential of MFA to identify cases of high or low emissions, the observations were re-classified into HIGH, MEDIUM and LOW according to the proportion of each individual MFA. The proportion of observations correctly classified were recorded. This was done for each individual MFA and for the calculated Gini coefficients, finding that a maximum of 67% of observations were correctly classified as HIGH CH4 (trans-12 C18:1) and a maximum of 58% of observations correctly classified as LOW CH4 (cis-9 C17:1). Gini coefficients did not improve this classification. These results suggest that MFA are not yet reliable predictors of specific amounts of CH4 emitted by a cow, while holding a modest potential to differentiate cases of high or low emissions.

Modelling three-dimensional fungal growth in response to environmental stimuli.

Most fungi grow by developing complex networks that enable the translocation of nutrients over large distances. Spatially explicit mathematical models are able to capture both the complexity of the fungal network and the biomass evolution, as such providing a powerful alternative to classical modelling paradigms. Unfortunately, most of these models restrict growth to two dimensions or confine it to a lattice, thereby resulting in unrealistic representations of fungal networks. In addition, interactions between fungi and their environment are often neglected. In response, this work presents a lattice-free three-dimensional fungal growth model that accounts for the interactions between the in silico fungus and different substrates and media. A sensitivity analysis was carried out to identify the key model parameters for future calibration. Finally, a scenario analysis covering a variety of growth conditions was conducted to illustrate the broad scope of the model and its ability to replicate in situ growth scenarios.

Separability Criteria for the Evaluation of Boundary Detection Benchmarks.

There exist a significant number of benchmarks for evaluating the performance of boundary detection algorithms, most of them relying on some sort of comparison of the automatically-generated boundaries with human-labeled ones. Such benchmarks are composed of a representative image data set, as well as a comparison measure on the universe of boundary images. Despite many such data sets and measures have been proposed, there is no clear way of knowing which combinations of them are the most suitable for the task. In this paper, we introduce four criteria that allow for a sensible evaluation of the performance of a comparison measure on a given data set. The criteria mimic the way in which humans understand boundary images, as well as their ability to recognize the underlying scenes. These criteria can, as a final goal, quantify the ability of the boundary detection benchmarks to evaluate the performance of boundary detection methods, either edge-based or segmentation-based.

RECIPE COMPLETION USING MACHINE LEARNING TECHNIQUES.

Completing a recipe is a non-trivial task, as the success of ingredient combinations depends on a multitude of factors such as taste, smell, texture, etc. The aim of our work is to build a model that adds one or more ingredients to a given number of ingredients. The idea is based on leftover ingredients in a fridge. A person could list the available ingredients in his or her fridge and the model would suggest some additional ingredients to create a full recipe.

Automated image-based analysis of spatio-temporal fungal dynamics.

Due to their ability to grow in complex environments, fungi play an important role in most ecosystems and have for that reason been the subject of numerous studies. Some of the main obstacles to the study of fungal growth are the heterogeneity of growth environments and the limited scope of laboratory experiments. Given the increasing availability of image capturing techniques, a new approach lies in image analysis. Most previous image analysis studies involve manual labelling of the fungal network, tracking of individual hyphae, or invasive techniques that do not allow for tracking the evolution of the entire fungal network. In response, this work presents a highly versatile tool combining image analysis and graph theory to monitor fungal growth through time and space for different fungal species and image resolutions. In addition, a new experimental set-up is presented that allows for a functional description of fungal growth dynamics and a quantitative mutual comparison of different growth behaviors. The presented method is completely automated and facilitates the extraction of the most studied fungal growth features such as the total length of the mycelium, the area of the mycelium and the fractal dimension. The compactness of the fungal network can also be monitored over time by computing measures such as the number of tips, the node degree and the number of nodes. Finally, the average growth angle and the internodal length can be extracted to study the morphology of the fungi. In summary, the introduced method offers an updated and broader alternative to classical and narrowly focused approaches, thus opening new avenues of investigation in the field of mycology.

Milk fatty acids as possible biomarkers to diagnose hyperketonemia in early lactation.

The aim of this study was to assess the potential of milk fatty acids as diagnostic tool for hyperketonemia of 93 dairy cows in a 3×2 factorial arrangement. Cows were fed a glucogenic or lipogenic diet and originally were intended to be subjected to a 0-, 30-, or 60-d dry period. Nevertheless, some of the cows, which were intended for inclusion in the 0-d dry period group, dried off spontaneously. Milk was collected in wk 2, 3, 4, and 8 of lactation for milk fat analysis. Blood was sampled from wk 2 to 8 after parturition for ß-hydroxybutyrate (BHBA) analysis. Cases were classified into 2 groups: hyperketonemia (BHBA ≥1.2mmol/L) and nonhyperketonemia (BHBA <1.2mmol/L). Concentrations of 45 milk fatty acids and ratios of anteiso C15:0-to-anteiso C17:0 and C18:1 cis-9-to-C15:0 were subjected to a logistic regression analysis (stepwise forward method). The milk fat C18:1 cis-9-to-C15:0 ratio revealed the most discriminating factor for diagnosis of hyperketonemia. Ninety percent of nonhyperketonemia cases showed a milk fat C18:1 cis-9-to-C15:0 ratio of 40 or lower, whereas 70% of cows suffering from hyperketonemia showed milk fat C18:1 cis-9-to-C15:0 ratios exceeding 40. Additionally, cows with a milk fat ratio C18:1 cis-9-to-C15:0 of at least 45 in wk 2 after parturition had about 50% chance to encounter blood plasma BHBA values of 1.2mmol/L or more during the first 8 wk of lactation. Of the cows not suffering from hyperketonemia during the first 2 mo of lactation, only 9% exceeded this wk 2 threshold. Practical implementation requires routine analysis of both milk fatty acids, which currently is lacking for C15:0. The inclusion of other variables, such as test-day information and a more frequent sampling protocol should be considered to further improve diagnostic performance of this biomarker.

A new protocol for evaluating the efficacy of some dispensing systems of a packaging in the microbial protection of water-based preservative-free cosmetic products.

OBJECTIVE: A new protocol is described for assessing the efficacy of the dispenser of some packaging systems (PSs) of preservative-free cosmetic products in protecting both their contained formula and their delivered doses. METHODS: Practically, aiming at mimicking contacts with a non-sterile skin or fingers, the dispensing system is put into contact with a pre-contaminated fabric by a standardized colonization of P. aeruginosa. RESULTS: When applied to three different types of packaging, results show clear differences in both criteria between these conditioning articles, that is variable efficacies in protecting the contained product and the delivered doses, knowing that the first aspect is of paramount importance. CONCLUSION: The proposed protocol is proved being able to discriminate between different PSs and provides information on strong and weak features of certain types dispensing technologies prone to efficiently decrease either the dose contamination or to prevent contamination in reaching the contained product. Therefore, the proposed protocol can contribute to an objective selection of a PS for protecting a cosmetic care product with a low content of preservative or preservative free.

Milk fatty acids as possible biomarkers to early diagnose elevated concentrations of blood plasma nonesterified fatty acids in dairy cows.

Most cows encounter a state of negative energy balance during the periparturient period, which may lead to metabolic disorders and impaired fertility. The aim of this study was to assess the potential of milk fatty acids as diagnostic tools of detrimental levels of blood plasma nonesterified fatty acids (NEFA), defined as NEFA concentrations beyond 0.6 mmol/L, in a data set of 92 early lactating cows fed a glucogenic or lipogenic diet and subjected to 0-, 30-, or 60-d dry period before parturition. Milk was collected in wk 2, 3, 4, and 8 (n = 368) and blood was sampled weekly from wk 2 to 8 after parturition. Milk was analyzed for milk fatty acids and blood plasma for NEFA. Data were classified as "at risk of detrimental blood plasma NEFA" (NEFA ≥ 0.6 mmol/L) and "not at risk of detrimental blood plasma NEFA" (NEFA <0.6 mmol/L). Concentrations of 45 milk fatty acids and milk fat C18:1 cis-9-to-C15:0 ratio were subjected to a discriminant analysis. Milk fat C18:1 cis-9 revealed the most discriminating variable to identify detrimental blood plasma NEFA. A false positive rate of 10% allowed us to diagnose 46% of the detrimental blood plasma NEFA cases based on a milk fat C18:1 cis-9 concentration of at least 230 g/kg of milk fatty acids. Additionally, it was assessed whether the milk fat C18:1 cis-9 concentrations of wk 2 could be used as an early warning for detrimental blood plasma NEFA risk during the first 8 wk in lactation. Cows with at least 240 g/kg of C18:1 cis-9 in milk fat had about 50% chance to encounter blood plasma NEFA values of 0.6 mmol/L or more during the first 8 wk of lactation, with a false positive rate of 11.4%. Profit simulations were based on costs for cows suffering from detrimental blood plasma NEFA, and costs for preventive treatment based on daily dosing of propylene glycol for 3 wk. Given the relatively low incidence rate (8% of all observations), continuous monitoring of milk fatty acids during the first 8 wk of lactation to diagnose detrimental blood plasma NEFA does not seem cost effective. On the contrary, milk fat C18:1 cis-9 of the second lactation week could be an early warning of cows at risk of detrimental blood NEFA. In this case, selective treatment may be cost effective.

Towards combinatorial spectroscopy: the case of minor milk fatty acids determination.

Chemometrical models for determination of milk fatty acids (FA) are typically developed using spectral data from a single spectroscopy technique, e.g., mid-infrared spectroscopy in milk control. Such models perform poorly in determining minor components and are highly dependent on the spectral data source and on the type of matrix. In milk fat, the unsuccessful determination of minor (fatty acids lower than 1.0 g/100g in total fat) FA is often the result of: (1) the molecular structure similarity between the minor and the major FA within the milk fat matrix (thus the chemical signature specific to individual fatty acids has restricted specificity), and (2) the low signal intensity (detection limit) for specific vibrational modes. To overcome these limitations, data from different types of spectroscopy techniques, which brings additional chemical information in relation to the variation of the FA, could be included in the regression models to improve quantification. Here, Fourier transform (FT) Raman spectra were concatenated with attenuated total reflectance FT infrared (ATR/FTIR) spectra. The new combinatorial models showed up to 25% decrease in the root mean squared error of cross-validation (RMSECV) values, accompanied with a higher Rcv(2) for most individual FA or sums of FA groups, as compared to regression models based on Raman only or ATR/FTIR only spectra. In addition, improved models included less PLS components indicating an increased robustness. Interpretation of the most contributing regression coefficients indicated the value of newly combined spectral regions as carriers of specific chemical information. Although requiring additional spectroscopy instrumentation and prolonged acquisition time, this new combinatorial approach can be automated and is sufficient for semi-routine determination of the milk FA profile.

Grain-based versus alfalfa-based subacute ruminal acidosis induction experiments: Similarities and differences between changes in milk fatty acids.

Subacute ruminal acidosis (SARA) is one of the most important metabolic disorders, traditionally characterized by low rumen pH, which might be induced by an increase in the dietary proportion of grains as well as by a reduction of structural fiber. Both approaches were used in earlier published experiments in which SARA was induced by replacing part of the ration by a grain mixture or alfalfa hay by alfalfa pellets. The main differences between both experiments were the presence of blood lipopolysaccharide and Escherichia coli and associated effects on the rumen microbial population in the rumen of grain-based induced SARA animals as well as a great amount of quickly fermentable carbohydrates in the grain-based SARA induction experiment. Both induction approaches changed rumen pH although the pH decrease was more substantial in the alfalfa-based SARA induction protocol. The goal of the current analysis was to assess whether both acidosis induction approaches provoked similar shifts in the milk fatty acid (FA) profile. Similar changes of the odd- and branched-chain FA and the C18 biohydrogenation intermediates were observed in the alfalfa-based SARA induction experiment and the grain-based SARA induction experiment, although they were more pronounced in the former. The proportion of trans-10 C18:1 in the last week of the alfalfa-based induction experiment was 6 times higher than the proportion measured during the control week. The main difference between both induction experiments under similar rumen pH changes was the decreasing sum of iso FA during the grain-based SARA induction experiment whereas the sum of iso FA remained stable during the alfalfa-based SARA induction experiment. The cellulolytic bacterial community seemed to be negatively affected by either the presence of E. coli and the associated lipopolysaccharide accumulation in the rumen or by the amount of starch and quickly fermentable carbohydrates in the diet. In general, changes in the milk FA profile were related to changes in rumen pH. Nevertheless, feed characteristics (low in structural fiber vs. high in starch) also affected the milk FA profile and, as such, both effects should be taken into account when subacute acidosis occurs.