Martin Høy

Regression of a data matrix on descriptors of both its rows and of its columns via latent variables: L-PLSR

Abstract A new approach is described, for extracting and visualising structures in a data matrix Y in light of additional information BOTH about the ROWS in Y, given in matrix X, AND about the COLUMNS in Y, given in matrix Z. The three matrices Z–Y–X may be envisioned as an “L-shape”; X(I×K) and Z(J×L) share no matrix size dimension, but are connected via Y(I×J). A few linear combinations (components) are extracted from X and from Z, and their interactions are used for bi-linear modelling of Y, as well as for bi-linear modelling of X and Z themselves. The components are defined by singular value decomposition (SVD) of X′YZ. Two versions of the L-PLSR are described—using one single SVD for all components, or component-wise SVDs after deflation. The method is applied to the analysis of consumer liking data Y of six products assessed by 125 persons, in light of 10 other product descriptors X and 15 other person descriptors Z. Its performance is also checked on artificial data.

Analysis of designed experiments by stabilised PLS Regression and jack-knifing

Abstract Pragmatical, visually oriented methods for assessing and optimising bi-linear regression models are described, and applied to PLS Regression (PLSR) analysis of multi-response data from controlled experiments. The paper outlines some ways to stabilise the PLSR method to extend its range of applicability to the analysis of effects in designed experiments. Two ways of passifying unreliable variables are shown. A method for estimating the reliability of the cross-validated prediction error RMSEP is demonstrated. Some recently developed jack-knifing extensions are illustrated, for estimating the reliability of the linear and bi-linear model parameter estimates. The paper illustrates how the obtained PLSR “significance” probabilities are similar to those from conventional factorial ANOVA, but the PLSR is shown to give important additional overview plots of the main relevant structures in the multi-response data. The study is part of an ongoing effort to establish a cognitively simple and versatile approach to multivariate data analysis, with reliability assessment based on the data at hand, and with little need for abstract distribution theory [H. Martens, M. Martens, Multivariate Analysis of Quality. An Introduction, Wiley, Chichester, UK, 2001].

Reduction of verotoxigenic Escherichia coli in production of fermented sausages

After a number of foodborne outbreaks of verotoxigenic Escherichia coli involving fermented sausages, some countries have imposed regulations on sausage production. For example, the US Food Safety and Inspection Service requires a 5 log(10) reduction of E. coli in fermented products. Such regulations have led to a number of studies on the inactivation of E. coli in fermented sausages by changing processing and post-processing conditions. Several factors influence the survival of E. coli such as pre-treatment of the meat, amount of NaCl, nitrite and lactic acid, water activity, pH, choice of starter cultures and addition of antimicrobial compounds. Also process variables like fermentation temperature and storage time play important roles. Though a large variety of different production processes of sausages exist, generally the reduction of E. coli caused by production is in the range 1-2 log(10). In many cases this may not be enough to ensure microbial food safety. By optimising ingredients and process parameters it is possible to increase E. coli reduction to some extent, but in some cases still other post process treatments may be required. Such treatments may be storage at ambient temperatures, specific heat treatments, high pressure processing or irradiation. HACCP analyses have identified the quality of the raw materials, low temperature in the batter when preparing the sausages and a rapid pH drop during fermentation as critical control points in sausage production. This review summarises the literature on the reduction verotoxigenic E. coli in production of fermented sausages.

REVIEW OF PARTIAL LEAST SQUARES REGRESSION PREDICTION ERROR IN UNSCRAMBLER

Abstract Three expressions for estimating the prediction uncertainty in partial least squares regression (PLSR) have been evaluated, using synthetic datasets and Monte Carlo simulations. The simulations revealed that the original expression used in the old Unscrambler program needed a correcting factor, as pointed out in a recently published article. With low noise levels, the corrected uncertainty estimator used in the latest version of Unscrambler (7.0) performed reasonably well as an estimator of the actual prediction error. A third estimate proposed in another recently published article seemed to lack a term to differentiate between the prediction objects, and thus did not perform satisfactorily.

Feasibility of NIR interactance hyperspectral imaging for on-line measurement of crude composition in vacuum packed dry-cured ham slices

There is a growing market for packaged slices of dry-cured ham. The heterogeneity of the composition of slices between packages is an important drawback when aiming to offer consumers a product with a known and constant composition which fits individual consumer expectations. The aim of this work was to test the feasibility of NIR interactance imaging for on-line analysis of water, fat and salt and their spatial distribution in dry-cured ham slices. PLSR models for predicting water, fat and salt contents with NIR spectra were developed with a calibration set of samples (n=82). The models were validated with an external validation set (n=42). The predictive models were accurate enough for screening purposes. The errors of prediction were 1.34%, 1.36% and 0.71% for water, fat and salt, respectively. The spatial distribution of these components within the slice was also obtained.

Use of computed tomography to study raw ham properties and predict salt content and distribution during dry-cured ham production

Varying salt content in hams of equal brand is a major challenge for Norwegian dry-cured ham producers. This study was thus undertaken to test existing computed tomography (CT) calibration models for salt on entire hams, regarding predictability of salt content at different processing times including final ham and to study salt distribution during processing of dry-cured ham. Twenty-six hams were scanned by computed tomography (CT) 11 times during dry-curing for this purpose. However, previously established calibration models had to be adjusted as they overestimated salt in dry samples. Prediction of ultimate salt content was more accurate approaching the end of the dry-curing process (RMSEP=0.351-0.595% salt). Inclusion of remaining weight loss improved the prediction accuracy in un-dried samples by approximately 0.1% NaCl. The prediction errors were sufficiently low to be of practical interest.

Proteome changes in the insoluble protein fraction of bovine Longissimus dorsi muscle as a result of low-voltage electrical stimulation

Changes induced by low-voltage electrical stimulation (ES; 0-95 V for 8 s; 95 V for 32 s) in the insoluble protein fraction of bovine longissimus dorsi (LD) muscle at 1 and 24h post-ES were investigated by proteomics. Protein abundance patterns from ten Norwegian Red (NRF) young bulls were compared, and significant changes due to ES were found by rotation test and partial least square (PLS) regression analyses. Five protein spots showed lower abundance in ES samples at both sampling times, and in addition, 10 proteins at 1 h post-ES and 13 proteins at 24 h post-ES changed significantly in abundance due to ES. Reduced abundance of full-length structural proteins in ES samples indicates an accelerated proteolysis due to ES. Moreover, increased abundance of small heat shock proteins indicates earlier initiation of stress responses due to ES. These findings provide a better understanding of the biochemical processes taking place as a result of ES during post mortem storage of meat.

Multi-level binary replacement (MBR) design for computer experiments in high-dimensional nonlinear systems

Computer experiments are useful for studying a complex system, e.g. a high‐dimensional nonlinear mathematical model of a biological or physical system. Based on the simulation results, an empirical “metamodel” may then be developed, emulating the behavior of the model in a way that is faster to compute and easier to understand. In modelometrics, the model phenome of a computer model is recorded, once and for all, by structured simulations according to a factorial design in the model inputs, and with high‐dimensional profiling of its simulation outputs. A multivariate metamodel is then developed, by multivariate analysis of the input–output data, akin to how high‐dimensional data are analyzed in chemometrics. To reveal strongly nonlinear input–output relationships, the factorial design must probe the design space at many different levels for each of the many input factors. A reduced factorial design method may be required if combinatorial explosion is to be avoided. In the multi‐level binary replacement (MBR) design the levels of each input factor are represented as binary numbers, and all the individual binary factor bits are then combined in a fractional factorial (FF) design. The experiment size can thereby be greatly reduced at the price of some binary confounding. The MBR method is here described and then illustrated for the optimization of a nonlinear model of a microbiological growth curve with five design factors, for finding the relevant region in the design space, and subsequently for estimating the optimal design points in that space. Copyright

Shiga toxigenic Escherichia coli show strain dependent reductions under dry-fermented sausage production and post-processing conditions.

Dry-fermented sausages (DFS) are considered possible risk products regarding Shiga toxigenic Escherichia coli (STEC). We have compared the reduction of 11 E. coli isolates of various serogroups in salami during the sausage production process and during post-process measures including storage, heating and freezing. The 11 E. coli isolates, mainly STEC, included enterohaemorrhagic E. coli (EHEC) outbreak strains linked to DFS along with apathogenic E. coli. During sausage production, there was a statistically significant difference in reduction between the E. coli strains ranging from 1.3 to 2.4 log₁₀ (p<0.001). When sausages were subjected to post-process heat treatment of 43 °C for 24 h, a total reduction of more than 5 log₁₀ was obtained for all E. coli isolates. Freezing and thawing of DFS with subsequent storage for 1 month at 16 or 20 °C generally contributed to large E. coli reductions with the latter conditions giving an average additional 3.9 log₁₀ reduction, with a range from 3.4 to 4.4 log₁₀. The combination of freezing and 1 month of storage gave higher reductions compared with storage for 2 months for all examined temperatures. No systematic differences in survival of E. coli of different serogroups were detected for the different post-process measures. The reductions were also similar to those of apathogenic control isolates. Isolates showing higher survival during the ripening process did not have a lower reduction when exposed to post-process stress like storage, heating and freezing. The ability of the isolates to survive in salami was also compared with their survival at equivalent conditions in a tryptic soy broth (TSB) model. There was a low and not significant correlation (p>0.1) between the reductions of E. coli in salami and in the TSB broth model. Results based on broth models and/or single or surrogate strains must therefore be interpreted with caution. The EHEC reducing post-processing measures tested can easily be implemented in DFS production with marginal influence on the quality of the sausages.

Effects of food ingredients and oxygen exposure on premature browning in cooked beef

Premature browning (PMB) in the centre of cooked hamburgers and beef loin (M. longissimus dorsi) steaks was assessed visually and instrumentally. Rosemary extract, ascorbic acid, sodium lactate, polyphosphate or lingonberry juice were added to freshly ground beef with predominant oxymyoglobin, and hamburgers were cooked to 62°C. In general, the tested ingredients did not reduce the extent of PMB in hamburgers, but polyphosphate tended to reduce PMB due to increased pH. Control burgers made of vacuum packaged meat with deoxymyoglobin were cooked to 62, 69 and 75°C, and did not express PMB. Beef loins were injected with a solution of sodium lactate, polyphosphate and sodium chloride. Loin steaks were stored under 75% O(2)/25% CO(2) for 5 days and also cooked to 62°C. Injected steaks had less PMB than non-injected controls, but of a low magnitude unlikely to influence the perception of doneness. The study demonstrated that anaerobic packaging is the most efficient measure to avoid PMB in beef.