Diacylglycerol O-acyltransferase (DGAT) isoforms play a role in peridroplet mitochondrial fatty acid metabolism in bovine liver.

Hepatocellular lipid accumulation characterizes fatty liver in dairy cows. Lipid droplets (LD), specialized organelles that store lipids and maintain cellular lipid homeostasis, are responsible for the ectopic storage of lipids associated with several metabolic disorders. In recent years, non-ruminant studies have reported that LD-mitochondria interactions play an important role in lipid metabolism. Due to the role of diacylglycerol acyltransferase isoforms (DGAT1 and DGAT2) in LD synthesis, we explored mechanisms of mitochondrial fatty acid transport in ketotic cows using liver biopsies and isolated primary hepatocytes. Compared with healthy cows, cows with fatty liver had massive accumulation of LD and high protein expression of the triglyceride (TAG) synthesis-related enzymes DGAT1 and DGAT2, LD synthesis-related proteins perilipin 2 (PLIN2) and perilipin 5 (PLIN5), and the mitochondrial fragmentation-related proteins dynamin-related protein 1 (DRP1) and fission 1 (FIS1). In contrast, factors associated with fatty acid oxidation, mitochondrial fusion and mitochondrial electron transport chain complex were lower compared with those in the healthy cows. In addition, transmission electron microscopy revealed significant contacts between LD-mitochondria in liver tissue from cows with fatty liver. Compared with isolated cytoplasmic mitochondria, expression of carnitine palmitoyl transferase 1A (CPT1A) and DRP1 was lower, but mitofusin 2 (MFN2) and mitochondrial electron transport chain complex was greater in isolated peridroplet mitochondria from hepatic tissue of cows with fatty liver. In vitro data indicated that exogenous free fatty acids (FFA) induced hepatocyte LD synthesis and mitochondrial dynamics consistent with in vivo results. Furthermore, DGAT2 inhibitor treatment attenuated the FFA-induced upregulation of PLIN2 and PLIN5 and rescued the impairment of mitochondrial dynamics. Inhibition of DGAT2 also restored mitochondrial membrane potential and reduced hepatocyte reactive oxygen species production. The present in vivo and in vitro results indicated there are functional differences among different types of mitochondria in the liver tissue of dairy cows with ketosis. Activity of DGAT2 may play a key role in maintaining liver mitochondrial function and lipid homeostasis in dairy cows during the transition period.

Microbial profile during fermentation and aerobic stability of ensiled mixtures of maize stover and banana pseudostem in South Ethiopia.

AIMS: This study evaluated pH reduction and microbial growth during fermentation of maize stover (MS) mixed with banana pseudostem (BPS) under South Ethiopian conditions. MATERIALS AND RESULTS: The MS and BPS were chopped and mixed into six treatments (T): 80% BPS plus 20% DMS (T1), 70% BPS plus 30% DMS (T2), 40% BPS plus 60% FMS (fresh MS) (T3), 20% BPS plus 80% FMS (T4), 100% FMS (T5), and 95% BPS plus 5% molasses (T6). At 0, 7, 14, 30, 60, and 90 days, pH and dry matter were determined. Microbiological quality was assessed using plate counts and Illumina MiSeq sequencing. On day 60 and 90, aerobic stability was investigated. The results showed a significant reduction in pH in all mixtures, except in T1 and T2. Lactic acid bacteria counts reached a maximum in all treatments within 14 days. Sequencing showed marked changes in dominant bacteria, such as Buttiauxella and Acinetobacter to Lactobacillus and Bifidobacterium. CONCLUSIONS: The fresh MS and BPS mixtures and fresh maize showed significant pH reduction and dominance of desirable microbial groups. SIGNIFICANCE AND IMPACT OF THE STUDY: The study enables year-round livestock feed supplementation to boost milk and meat production in South Ethiopia.

Bridging the gap between measurement-based and simulation-based metamodels for deriving bulk optical properties from spatially-resolved reflectance profiles: effect of illumination and detection geometry.

Non-invasive determination of the optical properties is essential for understanding the light propagation in biological tissues and developing optical techniques for quality detection. Simulation-based models provide flexibility in designing the search space, while measurement-based models can incorporate the unknown system responses. However, the interoperability between these two types of models is typically poor. In this research, the mismatches between measurements and simulations were explored by studying the influences from light source and the incident and detection angle on the diffuse reflectance profiles. After reducing the mismatches caused by the factors mentioned above, the simulated diffuse reflectance profiles matched well with the measurements, with R2 values above 0.99. Successively, metamodels linking the optical properties with the diffuse reflectance profiles were respectively built based on the measured and simulated profiles. The prediction performance of these metamodels was comparable, both obtaining R2 values above 0.96. Proper correction for these sources of mismatches between measurements and simulations thus allows to build a simulation-based metamodel with a wide range of desired optical properties that is applicable to different measurement configurations.

Mid-infrared spectroscopic analysis of raw milk to predict the blood nonesterified fatty acid concentrations in dairy cows.

In high-yielding dairy cattle, severe postpartum negative energy balance is often associated with metabolic and infectious disorders that negatively affect production, fertility, and welfare. Mobilization of adipose tissue associated with negative energy balance is reflected through an increased level of nonesterified fatty acids (NEFA) in the blood plasma. Earlier, identification of negative energy balance through detection of increased blood plasma NEFA concentration required laborious and stressful blood sampling. More recently, attempts have been made to predict blood NEFA concentration from milk samples. In this study, we aimed to develop and validate a model to predict blood plasma NEFA concentration using the milk mid-infrared (MIR) spectra that are routinely measured in the context of milk recording. To this end, blood plasma and milk samples were collected in wk 2, 3, and 20 postpartum for 192 lactations in 3 herds. The blood plasma samples were taken in the morning, and representative milk samples were collected during the morning and evening milk sessions on the same day. To predict plasma NEFA concentration from the milk MIR spectra, partial least squares regression models were trained on part of the observations from the first herd. The models were then thoroughly validated on all other observations of the first herd and on the observations of the 2 independent herds to explore their robustness and wide applicability. The final model could accurately predict blood plasma NEFA concentrations <0.6 mmol/L with a root mean square error of prediction of <0.143 mmol/L. However, for blood plasma with >1.2 mmol/L NEFA, the model clearly underestimated the true level. Additionally, we found that morning blood plasma NEFA levels were predicted with significantly higher accuracy using MIR spectra of evening milk samples compared with MIR spectra of morning samples, with root mean square error of prediction values of, respectively, 0.182 and 0.197 mmol/L, and R2 values of 0.613 and 0.502. These results suggest a time delay between variations in blood plasma NEFA and related milk biomarkers. Based on the MIR spectra of evening milk samples, cows at risk for negative energy status, indicated by detrimental morning blood plasma NEFA levels (>0.6 mmol/L), could be identified with a sensitivity and specificity of, respectively, 0.831 and 0.800. As this model can be applied to millions of historical and future milk MIR spectra, it opens an opportunity for regular metabolic screening and improved resilience phenotyping.

Estimation of Particle Size Distribution from Bulk Scattering Spectra: Validation on Monomodal Suspensions.

A particle size distribution (PSD) estimation method based on light-scattering properties was validated on experimental visible/near-infrared scattering spectra of polystyrene suspensions, with a nominal particle size ranging from 0.1 to 12 µm in diameter. On the basis of µs and g spectra extracted from double integrating sphere measurements, good PSD estimates were obtained for particles ≥1 µm. The particle volume fraction estimates in the case of µs were close to the target concentrations, although influenced by small baseline fluctuations on the spectra. For submicrometer particles, on the other hand, the non-oscillating µs spectra lack discriminating power, resulting in erroneous PSD estimates. The reduced scattering coefficient spectra (µs') were found less useful for particle size estimation as they lack a characteristic shape, causing an over- or underestimation of the distribution width. In summary, the estimation routine proved to deliver PSD estimates in line with the reference measurements for micrometer-sized or larger particles based on their µs and g scattering spectra. Additional validation on more polydisperse samples forms the next step before going to bimodal PSD estimates.

Short communication: Validation of a novel milk progesterone-based tool to monitor luteolysis in dairy cows using cost-effective, on-farm measured data.

The progesterone (P4) monitoring algorithm using synergistic control (PMASC) uses luteal dynamics to identify fertility events in dairy cows. This algorithm employs a combination of mathematical functions describing the increasing and decreasing P4 concentrations during the development and regression of the corpus luteum and a statistical control chart that allows identification of luteolysis. The mathematical model combines sigmoidal functions from which the cycle characteristics can be calculated. Both the moment at which luteolysis is detected and confirmed by PMASC, as well as the model features themselves, can be used to inform the farmer on the fertility status of the cows.

Validation of a novel milk progesterone-based tool to monitor luteolysis in dairy cows: Timing of the alerts and robustness against missing values.

Automated monitoring of fertility in dairy cows using milk progesterone is based on the accurate and timely identification of luteolysis. In this way, well-adapted insemination advice can be provided to the farmer to further optimize fertility management. To properly evaluate and compare the performance of new and existing data-processing algorithms, a test data set of progesterone time-series that fully covers the desired variability in progesterone profiles is needed. Further, the data should be measured with a high frequency to allow rapid onset events, such as luteolysis, to be precisely determined. Collecting this type of data would require a lot of time, effort, and budget. In the absence of such data, an alternative was developed using simulated progesterone profiles for multiple cows and lactations, in which the different fertility statuses were represented. To these, relevant variability in terms of cycle characteristics and measurement error was added, resulting in a large cost-efficient data set of well-controlled but highly variable and farm-representative profiles. Besides the progesterone profiles, information on (the timing of) luteolysis was extracted from the modeling approach and used as a reference for the evaluation and comparison of the algorithms. In this study, 2 progesterone monitoring tools were compared: a multiprocess Kalman filter combined with a fixed threshold on the smoothed progesterone values to detect luteolysis, and a progesterone monitoring algorithm using synergistic control, PMASC, which uses a mathematical model based on the luteal dynamics and a statistical control chart to detect luteolysis. The timing of the alerts and the robustness against missing values of both algorithms were investigated using 2 different sampling schemes: one sample per cow every 8 h versus 1 sample per day. The alerts for luteolysis of the PMASC algorithm were on average 20 h earlier compared with the ones of the multiprocess Kalman filter, and their timing was less sensitive to missing values. This was shown by the fact that, when 1 sample per day was used, the Kalman filter gave its alerts on average 24 h later, and the variability in timing of the alerts compared with simulated luteolysis increased with 22%. Accordingly, we postulate that implementation of the PMASC system could improve the consistency of luteolysis detection on farm and lower the analysis costs compared with the current state of the art.

Short communication: Sensitivity of estrus alerts and relationship with timing of the luteinizing hormone surge.

Both the sensitivity of an estrus detection system and the consistency of alarms relative to ovulation determine its value for a farmer. The objective of this study was to compare an activity-based system and a milk progesterone-based system for their ability to detect estrus reliably, and to investigate how their alerts are linked to the time of the LH surge preceding ovulation. The study was conducted on an experimental research farm in Flanders, Belgium. The activity alerts were generated by a commercial activity meter (ActoFIT, DeLaval, Tumba, Sweden), and milk progesterone was measured using a commercial ELISA kit. Sensitivity and positive predictive value of both systems were calculated based on 35 estrus periods over 43 d. Blood samples were taken for determination of the LH surge, and the intervals between timing of the alerts and the LH surge were investigated based on their range and standard deviation (SD). Activity alerts had a sensitivity of 80% and a positive predictive value of 65.9%. Alerts were detected from 39 h before until 8 h after the LH surge (range: 47 h, SD: 16 h). Alerts based on milk progesterone were obtained from a recently developed monitoring algorithm using a mathematical model and synergistic control. All estruses were correctly identified by this algorithm, and the LH surge followed, on average, 62 h later. Using the mathematical model, model-based indicators for the estimation of ovulation time can be calculated. Depending on which model-based indicator was used, ranges of 33 to 35 h and SD of about 11 h were obtained. Because detection of the LH surge was very labor intensive, only a limited number of potential estrus periods could be studied.

Estimation of particle size distributions from bulk scattering spectra: sensitivity to distribution type and spectral noise.

A shape dependent method for particle size distribution (PSD) estimation based on bulk scattering properties was elaborated. This method estimates the parameters of a particle size distribution with predefined shape from the bulk scattering spectra. The estimation routine was validated on simulated data of polystyrene in water suspensions. To investigate the effect of measurement errors on PSD estimates, a sensitivity analysis was performed. The influence of spectral resolution and range was rather limited. Good PSD estimations were obtained on noise-free spectra, spectra with limited random noise and for estimations on µs or µs' in case of a multiplicative baseline. However, the PSD estimation deteriorated if an incorrect value for the refractive index of the particle relative to the medium was used as input parameter. Deviations caused by an incorrect distribution type were smaller for more narrow PSDs than for broader ones. Overall, this study showed the potential to estimate PSDs from bulk scattering spectra and indicated the factors affecting the accuracy.

Method for short-term prediction of milk yield at the quarter level to improve udder health monitoring.

Udder health problems are often associated with milk losses. These losses are different between quarters, as infected quarters are affected both by systemic and pathogen-specific local effects, whereas noninfected quarters are only subject to systemic effects. To gain insight in these losses and the milk yield dynamics during disease, it is essential to have a reliable reference for quarter-level milk yield in an unperturbed state, mimicking its potential yield. We developed a novel methodology to predict this quarter milk yield per milking session, using an historical data set of 504 lactations collected on a test farm by an automated milking system from DeLaval (Tumba, Sweden). Using a linear mixed model framework in which covariates associated with the linearized Wood model and the milking interval are included, we were able to describe quarter-level yield per milking session with a proportional error below 10%. Applying this model enables us to predict the milk yield of individual quarters 1 to 50 d ahead with a mean prediction error ranging between 8 and 20%, depending on the amount of historical data available to estimate the random effect covariates for the predicted lactation. The developed methodology was illustrated using 2 examples for which quarter-level milk losses are calculated during clinical mastitis. These showed that the quarter-level mixed model allows us to gain insight in quarter lactation dynamics and enables to calculate milk losses in different situations.

A novel system for on-farm fertility monitoring based on milk progesterone.

Timely identification of a cow's reproduction status is essential to minimize fertility-related losses on dairy farms. This includes optimal estrus detection, pregnancy diagnosis, and the timely recognition of early embryonic death and ovarian problems. On-farm milk progesterone (P4) analysis can indicate all of these fertility events simultaneously. However, milk P4 measurements are subject to a large variability both in terms of measurement errors and absolute values between cycles. The objective of this paper is to present a newly developed methodology for detecting luteolysis preceding estrus and give an indication of its on-farm use. The innovative monitoring system presented is based on milk P4 using the principles of synergistic control. Instead of using filtering techniques and fixed thresholds, the present system employs an individually on-line updated model to describe the P4 profile, combined with a statistical process control chart to identify the cow's fertility status. The inputs for the latter are the residuals of the on-line updated model, corrected for the concentration-dependent variability that is typical for milk P4 measurements. To show its possible use, the system was validated on the P4 profiles of 38 dairy cows. The positive predictive value for luteolysis followed by estrus was 100%, meaning that the monitoring system picked up all estrous periods identified by the experts. Pregnancy or embryonic mortality was characterized by the absence or detection of luteolysis following an insemination, respectively. For 13 cows, no luteolysis was detected by the system within the 25 to 32 d after insemination, indicating pregnancy, which was confirmed later by rectal palpation. It was also shown that the system is able to cope with deviating P4 profiles having prolonged follicular or luteal phases, which may suggest the occurrence of cysts. Future research is recommended for optimizing sampling frequency, predicting the optimal insemination window, and establishing rules to detect problems based on deviating P4 patterns.

Anisotropic light propagation in bovine muscle tissue depends on the initial fiber orientation, muscle type and wavelength.

The effects of fiber orientation on vis/NIR light propagation were studied in three bovine muscles: biceps brachii, brachialis and soleus. Broadband light was focused onto the sample and the diffuse reflectance spot was captured using a hyperspectral camera (470-1620 nm), after which rhombuses were fitted to equi-intensity points. In samples with fibers running parallel to the measurement surface, the rhombus' major axis was oriented perpendicular to the fiber direction close to the point of illumination. However, at larger distances from the illumination spot, the major axis orientation aligned with the fiber direction. This phenomenon was found to be muscle dependent. Furthermore, the rhombus orientation was highly dependent on the sample positioning underneath the camera, especially when the muscle fibers ran parallel to the measurement surface. The bias parameter, indicating the deviation from a circular shape, was higher for samples with the fibers running parallel to the measurement surface. Moreover, clear effects of wavelength and distance from the illumination point on this parameter were observed. These results show the importance of fiber orientation when considering optical techniques for measurements on anisotropic, fibrous tissues. Moreover, the prediction of muscle fiber orientation seemed feasible, which can be of interest to the meat industry.

Three-dimensional microscale modelling of CO2 transport and light propagation in tomato leaves enlightens photosynthesis.

We present a combined three-dimensional (3-D) model of light propagation, CO2 diffusion and photosynthesis in tomato (Solanum lycopersicum L.) leaves. The model incorporates a geometrical representation of the actual leaf microstructure that we obtained with synchrotron radiation X-ray laminography, and was evaluated using measurements of gas exchange and leaf optical properties. The combination of the 3-D microstructure of leaf tissue and chloroplast movement induced by changes in light intensity affects the simulated CO2 transport within the leaf. The model predicts extensive reassimilation of CO2 produced by respiration and photorespiration. Simulations also suggest that carbonic anhydrase could enhance photosynthesis at low CO2 levels but had little impact on photosynthesis at high CO2 levels. The model confirms that scaling of photosynthetic capacity with absorbed light would improve efficiency of CO2 fixation in the leaf, especially at low light intensity.

Robust metamodel-based inverse estimation of bulk optical properties of turbid media from spatially resolved diffuse reflectance measurements.

Estimation of the bulk optical properties of turbid samples from spatially resolved reflectance measurements remains challenging, as the relation between the bulk optical properties and the acquired spatially resolved reflectance profiles is influenced by wavelength-dependent properties of the measurement system. The resulting measurement noise is apparent in the estimation of the bulk optical properties. In this study, a constrained inverse metamodeling approach is proposed to overcome these problems. First, a metamodel has been trained on a set of intralipid phantoms covering a wide range of optical properties to link the acquired spatially resolved reflectance profiles to the respective combinations of bulk optical properties (absorption coefficient and reduced scattering coefficient). In this metamodel, the wavelength (500 - 1700 nm) is considered as a third input parameter for the model to account for the wavelength dependent effects introduced by the measurement system. Secondly, a smoothness constraint on the reduced scattering coefficient spectra was implemented in the iterative inverse estimation procedure to robustify it against measurement noise and increase the reliability of the obtained bulk absorption and reduced scattering coefficient spectra. As the estimated values in some regions may be more reliable than others, the difference between simulated and measured values as a function of the evaluated absorption and scattering coefficients was combined in a 2D cost function. This cost function was used as a weight in the fitting procedure to find the parameters of the µ(s)' function giving the lowest cost over all the wavelengths together. In accordance with previous research, an exponential function was considered to represent the µ(s)' spectra of intralipid phantoms. The fitting procedure also provides an absorption coefficient spectrum which is in accordance with the measurements and the estimated parameters of the exponential function. This robust inverse estimation algorithm was validated on an independent set of intralipid® phantoms and its performance was also compared to that of a classical single-wavelength inverse estimation algorithm. While its performance in estimating µ(a) was comparable (R2 of 0.844 vs. 0.862), it resulted in a large improvement in the estimation of µ(s)' (R2 of 0.987 vs. 0.681). The change in performance is more apparent in the improvement of RMSE of µ(s)', which decreases from 10.36 cm(-1) to 2.10 cm(-1). The SRS profiles change more sensitively as a function of µ(a). As a result, there is a large range of µ(s)' and a small range of µa resulting in a good fit between measurement and simulation. The robust inverse estimator incorporates information over the different wavelengths, to increase the accuracy of µ(s)'estimations and robustify the estimation process.

Estimation of bulk optical properties of turbid media from hyperspectral scatter imaging measurements: metamodeling approach.

In many research areas and application domains, the bulk optical properties of biological materials are of great interest. Unfortunately, these properties cannot be obtained easily for complex turbid media. In this study, a metamodeling approach has been proposed and applied for the fast and accurate estimation of the bulk optical properties from contactless and non-destructive hyperspectral scatter imaging (HSI) measurements. A set of liquid optical phantoms, based on intralipid, methylene blue and water, were prepared and the Vis/NIR bulk optical properties were characterized with a double integrating sphere and unscattered transmittance setup. Accordingly, the phantoms were measured with the HSI technique and metamodels were constructed, relating the Vis/NIR reflectance images to the reference bulk optical properties of the samples. The independent inverse validation showed good prediction performance for the absorption coefficient and the reduced scattering coefficient, with R(2)(p) values of 0.980 and 0.998, and RMSE(P) values of 0.032 cm(-1) and 0.197 cm(-1) respectively. The results clearly support the potential of this approach for fast and accurate estimation of the bulk optical properties of turbid media from contactless HSI measurements.

Modeling the propagation of light in realistic tissue structures with MMC-fpf: a meshed Monte Carlo method with free phase function.

Monte Carlo methods commonly used in tissue optics are limited to a layered tissue geometry and thus provide only a very rough approximation for many complex media such as biological structures. To overcome these limitations, a Meshed Monte Carlo method with flexible phase function choice (fpf-MC) has been developed to function in a mesh. This algorithm can model the light propagation in any complexly shaped structure, by attributing optical properties to the different mesh elements. Furthermore, this code allows the use of different discretized phase functions for each tissue type, which can be simulated from the microstructural properties of the tissue, in combination with a tool for simulating the bulk optical properties of polydisperse suspensions. As a result, the scattering properties of tissues can be estimated from information on the microstructural properties of the tissue. This is important for the estimation of the bulk optical properties that can be used for the light propagation model, since many types of tissue have never been characterized in literature. The combination of these contributions, made it possible to use the MMC-fpf for modeling the light porapagation in plant tissue. The developed Meshed Monte Carlo code with flexible phase function choice (MMC-fpf) was successfully validated in simulation through comparison with the Monte Carlo code in Multi-Layered tissues (R2 > 0.9999) and experimentally by comparing the measured and simulated reflectance (RMSE = 0.015%) and transmittance (RMSE = 0.0815%) values for tomato leaves.

Dependent scattering in Intralipid® phantoms in the 600-1850 nm range.

The effect of dependent scattering on the bulk scattering properties of intralipid phantoms in the 600-1850 nm wavelength range has been investigated. A set of 57 liquid optical phantoms, covering a wide range of intralipid concentrations (1-100% v/v), was prepared and the bulk optical properties were accurately determined. The bulk scattering coefficient as a function of the particle density could be well described with Twersky's packing factor (R(2) > 0.990). A general model was elaborated taking into account the wavelength dependency and the effect of the concentration of scattering particles (R(2) = 0.999). Additionally, an empirical approach was followed to characterize the effect of dense packing of scattering particles on the anisotropy factor (R(2) = 0.992) and the reduced scattering coefficient (R(2) = 0.999) of the phantoms. The derived equations can be consulted in future research for the calculation of the bulk scattering properties of intralipid dilutions in the 600-1850 nm range, or for the validation of theories that describe the effects of dependent scattering on the scattering properties of intralipid-like systems.

Flexible tool for simulating the bulk optical properties of polydisperse spherical particles in an absorbing host: experimental validation.

In this study, a flexible tool to simulate the bulk optical properties of polydisperse spherical particles in an absorbing host medium is described. The generalized Mie solution for Maxwell's equations is consulted to simulate the optical properties for a spherical particle in an absorbing host, while polydispersity of the particle systems is supported by discretization of the provided particle size distributions. The number of intervals is optimized automatically in an efficient iterative procedure. The developed tool is validated by simulating the bulk optical properties for two aqueous nanoparticle systems and an oil-in-water emulsion in the visible and near-infrared wavelength range, taking into account the representative particle sizes and refractive indices. The simulated bulk optical properties matched closely (R2 ≥ 0.899) with those obtained by reference measurements.

On-farm NIR sensor for milk analysis: Exploitation of bias monitoring and bias correction.

Long-term studies have shown a bias drift over time in the prediction performance of near-infrared spectroscopy measurement systems. This bias drift generally requires extra laboratory reference measurements to detect and correct for this bias. Since these reference measurements are expensive and time consuming, there is a need for advanced methodologies for bias drift monitoring and correction without the need for taking extra samples. In this study, we propose and validate a method to monitor the bias drift and two methods to tackle it. The first method requires no extra measurements and uses a modified version of Partial Least Squares Regression to estimate and correct the bias. This method is based on the assumption that the mean concentration of the predicted component remains constant over time. The second method uses regular bulk milk measurements as a reference for bias correction. This method compares the measured concentrations of the bulk milk to the volume-weighted average concentrations of individual milk samples predicted by the sensor. Any difference between the actual and calculated bulk milk composition is then used to perform a bias correction on the predictions by the sensor system. The effectiveness of these methods to improve the component prediction was evaluated on data originating from a custom-built sensor that automatically measures the NIR reflectance and transmittance spectra of raw milk on the farm. We evaluate the practical use case where models for predicting the milk composition are trained upon installation of the sensor at the farm, and later used to predict the composition of subsequent samples over a period of more than 6 months. The effectiveness of the fully unsupervised method was confirmed when the mean concentration of the milk samples remained constant, while the effectiveness reduced when this was not the case. The bulk milk correction method was effective when all relevant samples for the component were measured by the sensor and included in the analyzed bulk milk, but is less effective when samples included in the bulk which are not measured by the sensor system. When the necessary conditions are met, these methods can be used to extend the lifetime of deployed prediction models by significantly reducing the bias on the predicted values.

PROSAC as a selection tool for SO-PLS regression: A strategy for multi-block data fusion.

BACKGROUND: Spectral data from multiple sources can be integrated into multi-block fusion chemometric models, such as sequentially orthogonalized partial-least squares (SO-PLS), to improve the prediction of sample quality features. Pre-processing techniques are often applied to mitigate extraneous variability, unrelated to the response variables. However, the selection of suitable pre-processing methods and identification of informative data blocks becomes increasingly complex and time-consuming when dealing with a large number of blocks. The problem addressed in this work is the efficient pre-processing, selection, and ordering of data blocks for targeted applications in SO-PLS. RESULTS: We introduce the PROSAC-SO-PLS methodology, which employs pre-processing ensembles with response-oriented sequential alternation calibration (PROSAC). This approach identifies the best pre-processed data blocks and their sequential order for specific SO-PLS applications. The method uses a stepwise forward selection strategy, facilitated by the rapid Gram-Schmidt process, to prioritize blocks based on their effectiveness in minimizing prediction error, as indicated by the lowest prediction residuals. To validate the efficacy of our approach, we showcase the outcomes of three empirical near-infrared (NIR) datasets. Comparative analyses were performed against partial-least-squares (PLS) regressions on single-block pre-processed datasets and a methodology relying solely on PROSAC. The PROSAC-SO-PLS approach consistently outperformed these methods, yielding significantly lower prediction errors. This has been evidenced by a reduction in the root-mean-squared error of prediction (RMSEP) ranging from 5 to 25 % across seven out of the eight response variables analyzed. SIGNIFICANCE: The PROSAC-SO-PLS methodology offers a versatile and efficient technique for ensemble pre-processing in NIR data modeling. It enables the use of SO-PLS minimizing concerns about pre-processing sequence or block order and effectively manages a large number of data blocks. This innovation significantly streamlines the data pre-processing and model-building processes, enhancing the accuracy and efficiency of chemometric models.