Øyvind Langsrud

Explorative Multifactor Approach for Investigating Global Survival Mechanisms of Campylobacter jejuni under Environmental Conditions

ABSTRACT Explorative approaches such as DNA microarray experiments are becoming increasingly important in microbial research. Despite these major technical advancements, approaches to study multifactor experiments are still lacking. We have addressed this problem by using rotation testing and a novel multivariate analysis of variance (MANOVA) approach (50-50 MANOVA) to investigate interacting experimental factors in a complex experimental design. Furthermore, a new rotation testing based method was introduced to calculate false-discovery rates for each response. This novel analytical concept was used to investigate global survival mechanisms in the environment of the major food-borne pathogen C. jejuni. We simulated nongrowth environmental conditions by investigating combinations of the factors temperature (5 and 25°C) and oxygen tension (anaerobic, microaerobic, and aerobic). Data were generated with DNA microarrays for information about gene expression patterns and Fourier transform infrared (FT-IR) spectroscopy to study global macromolecular changes in the cell. Microarray analyses showed that most genes were either unchanged or down regulated compared to the reference (day 0) for the conditions tested and that the 25°C anaerobic condition gave the most distinct expression pattern with the fewest genes expressed. The few up-regulated genes were generally stress related and/or related to the cell envelope. We found, using FT-IR spectroscopy, that the amount of polysaccharides and oligosaccharides increased under the nongrowth survival conditions. Potential mechanisms for survival could be to down regulate most functions to save energy and to produce polysaccharides and oligosaccharides for protection against harsh environments. Basic knowledge about the survival mechanisms is of fundamental importance in preventing transmission of this bacterium through the food chain.

Fixed or random assessors in sensory profiling

This paper presents a discussion of the difference between fixed and mixed ANOVA models in sensory analysis. The two models are related to different ways of considering the assessors. Tests for hypotheses in the two models are presented. Alternative analysis techniques and alternative hypotheses are proposed. A simulation study of the effect of a common violation of standard ANOVA assumptions is presented in the appendix.

Analyzing Designed Experiments with Multiple Responses

Abstract This paper is an overview of a unified framework for analyzing designed experiments with univariate or multivariate responses. Both categorical and continuous design variables are considered. To handle unbalanced data, we introduce the so-called Type II* sums of squares. This means that the results are independent of the scale chosen for continuous design variables. Furthermore, it does not matter whether two-level variables are coded as categorical or continuous. Overall testing of all responses is done by 50-50 MANOVA, which handles several highly correlated responses. Univariate p-values for each response are adjusted by using rotation testing. To illustrate multivariate effects, mean values and mean predictions are illustrated in a principal component score plot or directly as curves. For the unbalanced cases, we introduce a new variant of adjusted means, which are independent to the coding of two-level variables. The methodology is exemplified by case studies from cheese and fish pudding production.

Autofluorescence quantifies collagen in sausage batters with a large variation in myoglobin content.

In order to clarify the potential of the method of autofluorescence for determining the collagen content in meat batter, an experimental design was used where the emission originating from collagen was unrelated to the effect of myoglobin absorbance on the emission spectra. Muscles like beef Masseter, beef Latissimus dorsi and pork Glutens medius assured a large variation in myoglobin content, and made absorbance from myoglobin account for 65-84% of the variation in the emission spectra at wavelengths normally thought interesting for collagen quantification. Collagen (range 1.3-4.0%) accounted for 7-18% of the variation in the autofluorescence emission spectra. Collagen content could be predicted from the emission spectra provided multivariate regression techniques were used. Pre-processing of spectra reduced the prediction error for collagen by 0.03-0.12%, depending on method used; the lowest prediction error obtained being 0.48%. The method of autofluorescence gave lower prediction errors for collagen content than did the method of near infrared reflectance when applied to the same batters.

Variation in the response to manipulation of post-mortem glycolysis in beef muscles by low-voltage electrical stimulation and conditioning temperature.

The aim of this study was to investigate how manipulation of glycolytic rate by post-mortem processing conditions influences quality of aged beef of two bovine muscles of different physiological character, longissimus dorsi (LD) and adductor (AD). Post-mortem glycolysis was manipulated by low-voltage electrical stimulation (LV-ES) of half carcasses and by chilling rate of the muscles. Multivariate statistical analysis was used to visualise the data, while ANOVA was used to identify significant effects and interactions. As expected there was a significant effect of LV-ES on the pH decline in the first hours post-mortem in both muscles. Moreover, significant effects of LV-ES on WB shear force measured 2 and 8 days after slaughter were observed for LD at both chilling temperatures, while for AD no effect on WB shear force was observed. Furthermore, the results revealed a large individual variation in the response of LV-ES on both pH decline and WB shear force, and this variation did not always correlate for the two responses. Some animals showed no response of LV-ES on pH decline, but still had an improved WB shear force, and vice versa. The results from this study indicate that there probably are other mechanisms than accelerated pH decline and prevention of cold-shortening, by which LV-ES can affect meat tenderness.

A unified framework for significance testing in fractional fractorials

The problem of finding significant effects in fractional factorial experiments is viewed as a variable selection problem. In this study, various forward selection and backward elimination strategies are compared. The smallest effects are pooled for error, and p-values are found by Monte Carlo simulations. Some attention is given to the question how the methods for subset selection in orthogonal designs can be generalized to more general linear regression models. The generalized methodology is applied to develop the so-called realistic model strategy. This new procedure for factorial designs makes use of the interaction hierarchy to increase power of the tests.

Optimised score plot by principal components of predictions

A common problem in statistics/chemometrics is to relate two data matrices (X and Y) to each other, with the purpose of either prediction or interpretation. Usually, one is interested in understanding which directions in Y-space that can be predicted by which directions in X-space. Several methods exist for this, for instance, PLS regression and canonical correlation. The present paper presents a new plot for visualising the relationship between X and Y. The plot is based on a decomposition of the X-space that is optimal with respect to Y-variance. The new procedure can accompany any regression method.

Restricted randomization and multiple responses in industrial experiments

Two issues regarding designed experiments are discussed; restrictions on randomization and multiple responses. The former is typically related to hard-to-vary factors and factors appearing in separate stages of a process experiment. Randomization restrictions should be taken into account in the construction of the design as well as in the statistical modelling. In the paper, a case study of sausage production is discussed, having a split-plot model with correlated multiple responses. Multiple responses are handled in two ways, by principal component analysis (PCA) followed by ANOVA of the principal components, and by a newly developed alternative, the ‘50–50 MANOVA’. Multiple tests of correlated response variables are also described. Practical aspects of the planning, performing, response measurements, and statistical analysis are emphasized throughout. Hence, the paper aims to extend the utility of statistical methods in industry by linking design of experiments to multivariate analysis of the responses. Copyright

Compositional Factors that Influence Lipid Peroxidation in Beef Juice and Standard Sausages.

In order to identify how different additives influenced lipid peroxidation formation, a sausage only using beef juice as pigment source and a standard beef-pork meat sausage were studied. The effects of different additives, including fish oil, myoglobin, nitrite, clove extract, and calcium sources on oxidation and sensory properties were examined. Both sausage systems were stored in 3 different manners prior to testing: (1) frozen immediately at -80 °C; (2) chilled stored for 2.5 weeks followed by fluorescent light illumination at 4 °C for another 2 wk; (3) frozen at -20 °C for 5 mo. The frozen group 3 showed the highest peroxide formation and thiobarbituric acid reactive substances (TBARS) for both sausage systems. Unpolar peroxides dominated in both systems. The clove extract could offset the peroxide formation from myoglobin/beef juice and/or fish oil, but the addition of clove flavor was recognized by the sensory panelists. Calcium addition reduced lipid peroxide formation. Added nitrite and fish oil seemed to interact to stimulate nitroso-myoglobin formation. Nitrite was identified to interact with clove addition and thereby, relatively speaking, increased TBARS. The 2 sausage systems generally ranked the additives similarly as pro- and antioxidants.

Explaining Correlations by Plotting Orthogonal Contrasts

This article describes a new plot that aids understanding the relationship between two response variables in a designed experiment. In addition to plotting the observed values directly, we make a scatterplot of orthogonal contrasts from the general linear model. This plot contains the same correlation information as the ordinary scatterplot. Therefore, one can interpret how the effects of the various design variables contribute to the correlation coefficient. This idea is also useful in more general cases. Any graphic presentation of the original observations can be accompanied by a corresponding plot of orthogonal contrasts that often will clarify the interpretation.