Analysis of feeding preference experiments

Summary Published studies of consumer feeding preferences using foods that experience autogenic change in mass, numbers, area, etc., on the time scale of a feeding trial fail to employ appropriate statistical analyses to incorporate controls for those food changes occurring in the absence of the consumer. The studies that run controls typically use them to calculate a constant correction factor, which is subtracted prior to formal data analysis. This procedure constitutes a non-rigorous suppression of variance that overstates the statistical significance of observed differences. The appropriate statistical analysis for preference tests with two foods is usually a simple t-test performed on the between-food differences in loss of mass (or numbers, area, etc.) comparing the results of experimentals with consumers to controls without consumers. Application of this recommended test procedure to an actual data set illustrates how low replication in controls, which is typical of most studies of feeding preference, inhibits detection of an apparently large influence of previous mechanical damage (simulated grazing) in reducing the attractiveness of a brown alga to a sea urchin.     

Design and analysis of multiple-choice feeding-preference experiments

Summary A serious omission in ecological methodology is the absence of a rigorous statistical procedure to analyse multiple-choice feeding-preference experiments. A sample of 21 studies in the littoral marine context shows that results from such experiments are used to study a variety of conceptual issues, ranging from nutritional biology to ecosystem dynamics. A majority of such studies have been incorrectly analysed. The analytical problem has two facets: (1) lack of independence in the simultaneous offer of food types and (2) the existence of autogenic changes particular to each food type. Problem (2) requires the use of control arenas without the consumer. A recent advance allows the rigorous analysis of experiments with two food types offered simultaneously. Here I propose a method for the multiple-choice case. For the first problem I suggest the use of multivariate statistical analysis, providing both a parametric and a nonparametric procedure. The second problem is solved using basic statistical theory. I analyse data from an experiment with the sea urchin Tetrapygus niger feeding on three species of algae: Ulva nematoidea, Gymnogongrus furcellatus, and Macrocystis pyrifera. The parametric and nonparametric procedures yielded similar results, and showed that when offered the three species of algae T. niger does not feed at random but shows a preference for U. nematoidea. The method requires that the number of replicates in the treatment and control arenas be the same, and greater than the number of food types. The method is useful for other kinds of multiple-choice experiments.     

Comments on design and analysis of multiple-choice feeding-preference experiments

An approach to the analysis of multiple-choice food selection experiments proposed recently is criticised on three grounds and modified methods are suggested.     

A note on the preference of naïve horses for different water bowls

The aim of this study was to determine which of four different models of automatic waterbowls naïve horses prefer. Two groups of horses (n = 8 and 11, respectively) and three individual horses were used to determine the percentage of daily water consumption from two float valve (Lister SB NT 100, Little Giant) and two push valve (Lister SB-1, Suevia 10P) waterbowls. The waterbowls were mounted on a panel 1.1 m above ground and spaced 0.8 m apart. Waterbowl positions were rotated every 48 h until each had been at all locations. Almost no consumption occurred from the push valve waterbowls during any trial. Though there was no significant difference between the two float valve waterbowls during the first group trial, there was a trend suggesting a preference (P = 0.060) for the SB NT 100 over the Little Giant in the second group trial. Because the horses came to drink as a group, individual preference trials were also conducted. The percentage of daily consumption was significantly greater (P = 0.001) from the SB NT 100 compared to the Little Giant in the individual trials. The larger, more open bowl, the greater volume of water that was maintained in the bowl due to the float valve, and the lack of a tendency to startle horses when the bowl refilled were most likely the reasons the Lister SB NT 100 was preferred.     

Gene expression analysis with the parametric bootstrap

Recent developments in microarray technology make it possible to capture the gene expression profiles for thousands of genes at once. With this data researchers are tackling problems ranging from the identification of 'cancer genes' to the formidable task of adding functional annotations to our rapidly growing gene databases. Specific research questions suggest patterns of gene expression that are interesting and informative: for instance, genes with large variance or groups of genes that are highly correlated. Cluster analysis and related techniques are proving to be very useful. However, such exploratory methods alone do not provide the opportunity to engage in statistical inference. Given the high dimensionality (thousands of genes) and small sample sizes (often <30) encountered in these datasets, an honest assessment of sampling variability is crucial and can prevent the over-interpretation of spurious results. We describe a statistical framework that encompasses many of the analytical goals in gene expression analysis; our framework is completely compatible with many of the current approaches and, in fact, can increase their utility. We propose the use of a deterministic rule, applied to the parameters of the gene expression distribution, to select a target subset of genes that are of biological interest. In addition to subset membership, the target subset can include information about relationships between genes, such as clustering. This target subset presents an interesting parameter that we can estimate by applying the rule to the sample statistics of microarray data. The parametric bootstrap, based on a multivariate normal model, is used to estimate the distribution of these estimated subsets and relevant summary measures of this sampling distribution are proposed. We focus on rules that operate on the mean and covariance. Using Bernstein's Inequality, we obtain consistency of the subset estimates, under the assumption that the sample size converges faster to infinity than the logarithm of the number of genes. We also provide a conservative sample size formula guaranteeing that the sample mean and sample covariance matrix are uniformly within a distance epsilon > 0 of the population mean and covariance. The practical performance of the method using a cluster-based subset rule is illustrated with a simulation study. The method is illustrated with an analysis of a publicly available leukemia data set.     

Comparison of fatty acids of marine fungi using multivariate statistical analysis

Summary Ten obligate marine fungi have as their principal fatty acids 160, 180, 181n9 and 182n6. The fatty acids ranged from 14 to 22 carbons, completely dominated by those with even numbers of carbons. The amount of unsaturated fatty acids varied between 35% and 80%. Each isolate contained small amounts of the acids 183n3 and 204n6. Branched, hydroxy- or cyclic fatty acids were not detected. Multivariate statistical, i.e. principal component analysis, showed that all ten strains could be distinguished on the basis of their fatty acid composition. These results indicate that the marine fungi do not have an unusual fatty acid composition and suggest that chemometric, multivariate analysis might be employed to confirm taxonomic relationships among these organisms.     

Incipient sexual isolation in the nasuta-albomicans complex of Drosophila: mating preference in male-, female- and multiple-choice mating experiments

Interracial divergence is an important facet of speciation. Thenasuta-albomicans complex ofDrosophila with sixteen morphologically identical, karyotypically different but cross-fertile races is an excellent system to study a few dimensions of raciation.Drosophila nasuta nasuta, Drosophila nasuta albomicans, Cytorace 1, Cytorace 2, Cytorace 3 and Cytorace 4 of this subgroup have been subjected to male-, female- and multiple-choice mating experiments. Out of 8456 crosses conducted, 7185 had successful matings. The overall impression is that mating is far from random amongst these six closely related races of thenasuta-albomicans complex. The males ofD. n. albomicans, Cytorace 1 and Cytorace 4 in male-choice, the females of Cytorace 1 and Cytorace 2 in female-choice, and the males and females ofD. n. nasuta, D. n. albomicans, Cytorace 1 and Cytorace 4 against the males and females of Cytorace 2 in multiple-choice experiments, had significantly more homogamic matings than expected. Thus in this study of evolutionary experimentation on raciation under laboratory conditions, we have documented the initiation of preference for con-specific matings among closely related and independently evolving members of thenasuta-albomicans complex ofDrosophila.     

A general framework for the statistical exploration of the ecological niche

We propose a new statistical framework for the exploratory analysis of the ecological niche, the “General niche-environment system factor analysis” (GNESFA). The data required for this analysis are (i) a table giving the values of the environmental variables in each environment unit (EU, e.g., the patches of habitat on a vector map), (ii) a set of weights measuring the availability of the EUs to the species (e.g., the proportion of the study area covered by a given patch), and (iii) a set of utilization weights describing the use of the EUs by the focal species (e.g., the proportion of detections of the species in each patch). Each row of the table corresponds to a point in the multidimensional space defined by the environmental variables, and each point is associated with two weights. The GNESFA searches the directions in this space where the two weight distributions differ the most, choosing one distribution as the reference, and the other one as the focus. The choice of the utilization as the reference corresponds to the MADIFA (Mahalanobis distances factor analysis), which identifies the directions on which the available EUs are in average the furthest from the optimum of the niche, allowing habitat suitability modelling. The choice of the availability as the reference corresponds to the FANTER (Factor analysis of the niche, taking the environment as the reference), which identifies the directions on which the niche is the furthest from the average environment (marginality) and those on which the niche is the narrowest compared with the environment (specialization). The commonly used ENFA (Ecological niche factor analysis) is at the middle point between the MADIFA and the FANTER, considering both distributions as the reference and the focus simultaneously. When used concurrently, these three analyses allow an extensive exploration of the system.     

On the statistical significance of protein complex

Background: Statistical validation of predicted complexes is a fundamental issue in proteomics and bioinformatics. The target is to measure the statistical significance of each predicted complex in terms of p-values. Surprisingly, this issue has not received much attention in the literature. To our knowledge, only a few research efforts have been made towards this direction. Methods: In this article, we propose a novel method for calculating the p-value of a predicted complex. The null hypothesis is that there is no difference between the number of edges in target protein complex and that in the random null model. In addition, we assume that a true protein complex must be a connected subgraph. Based on this null hypothesis, we present an algorithm to compute the p-value of a given predicted complex. Results: We test our method on five benchmark data sets to evaluate its effectiveness. Conclusions: The experimental results show that our method is superior to the state-of-the-art algorithms on assessing the statistical significance of candidate protein complexes.     

Comparative proteomic analysis of planktonic and immobilized Pseudomonas aeruginosa cells: a multivariate statistical approach

The protein maps of Pseudomonas aeruginosa cells from two natural (attached) and one artificial (gel-entrapped) immobilized-cell (IC) systems, together with their free (suspended) counterparts, were compared after incubation for 18 or 48 h in a minimal salt medium. Principal component analysis (PCA) was used to interpret the variations in protein spot densities that were observed on electropherogram obtained by two-dimensional electrophoresis (2-DE). PCA of the 2-DE data, a matrix of 933 rows (observations, i.e., spot density values) and 12 columns (variables, i.e., incubation conditions), in which observations were standardized horizontally, extracted four principal components (PCs) accounting for 78.75% of the variability in the protein expression profiles. PC1 opposed the two modes of growth (planktonic and immobilized) while PC2 discriminated between the incubation times of free cell cultures. The incubation conditions of ICs, including the immobilization procedure (entrapment vs attachment) and the nature of the biofilm substratum, were fairly separated in PC3xPC4. The dependence of the protein patterns on the cell immobilization process was further illustrated by the identification of a number of peptides whose amount remained unchanged or was altered in ICs compared to free bacteria. These results reinforce the topical assertion that bacteria in the immobilized state display a specific physiological behavior but also question the existence of a unique IC phenotype.     

Integrating biological information into the statistical analysis and design of microarray experiments

Microarray technology is a powerful tool for animal functional genomics studies, with applications spanning from gene identification and mapping, to function and control of gene expression. Microarray assays, however, are complex and costly, and hence generally performed with relatively small number of animals. Nevertheless, they generate data sets of unprecedented complexity and dimensionality. Therefore, such trials require careful planning and experimental design, in addition to tailored statistical and computational tools for their appropriate data mining. In this review, we discuss experimental design and data analysis strategies, which incorporate prior genomic and biological knowledge, such as genotypes and gene function and pathway membership. We focus the discussion on the design of genetical genomics studies, and on significance testing for detection of differential expression. It is shown that the use of prior biological information can improve the efficiency of microarray experiments.     

A renewed approach to the nonparametric analysis of replicated microarray experiments

DNA-microarrays find broad employment in biochemical research. This technology allows the monitoring of the expression levels of thousands of genes at the same time. Often, the goal of a microarray study is to find differentially expressed genes in two different types of tissue, for example normal and cancerous. Multiple hypothesis testing is a useful statistical tool for such studies. One approach using multiple hypothesis testing is nonparametric analysis for replicated microarray experiments. In this paper we present an improved version of this method. We also show how p-values are calculated for all significant genes detected with this testing procedure. All algorithms were implemented in an R-package, and instructions on it's use are included. The package can be downloaded at http://www.statistik.unidortmund.de/de/content/einrichtungen/lehrstuehle/personen/jung.html     

Characterization of the caryopsis of common wheat varieties and F2 generation hybrids by means of multivariate analysis

Summary Nineteen parameters of common wheat kernels were investigated by the multivariate procedure. A relationship between the DBC parameter and both specific gravity and kernel filling was found. In some populations an interdependence of the DBC parameter and the subaleurone endosperm thickness was found. In the analysis of principal components the DBC parameter was represented mostly by particular component. Significant dependences were found between the dimensions as well as the weight of the caryopsis and some dimensions of the crease and endosperm cavity. The dendrite constructed on the basis of the calculated Mahalanobis's generalized distances matrix represents the near affinity of reciprocal hybrids or pure lines originating from a single cultivar as well as hybrids and their parental forms.     

Assessing the impact of plant invasions on soil seed bank communities: use of univariate and multivariate statistical approaches

Questions: Are soil seed banks affected by invasions of alien plants? How can we rigorously assess alterations in seed bank communities associated with invasive species and account for the high spatial variability of seed bank data? How do multivariate approaches compare with more traditional approaches based on analysis of variance? Location: Three riparian sites, Ireland. Methods: A protocol based on a combination of multivariate techniques was used to characterize soil seed bank communities associated with the herbaceous invasive species Heracleum mantegazzianum in May and October. Permutational multivariate analysis of variance (PERMANOVA) was used to test the effects of the factors “invasion”, “site”, “plot” and “depth” on the soil seed bank, while multivariate analysis of dispersion (PERMDISP) provided a measure of the variability of seed bank data at different spatial scales. Similarity percentages analysis (SIMPER) was used to identify the species that contributed most to the differences between invaded and uninvaded communities. A comparison between the results of PERMANOVA and ANOVA analyses was also made. Results: The composition of seed bank communities invaded by H. mantegazzianum differed significantly from that of uninvaded seed banks. Invaded seed banks were less diverse and had reduced abundance, and were dominated by only a few species, such as Urtica dioica and Juncus effusus. Such patterns were recorded at each of three depth categories, indicating that invasive plants can affect both the transient and the more persistent component of the soil seed bank. Seed bank variability was significantly higher within uninvaded areas, supporting the notion that invasions tend to lead to more homogeneous communities. Conclusion: The analytical protocol used in this study was effective in quantifying the effect of plant invasions, at different spatial scales, providing a statistically robust analysis of alterations in soil seed bank communities. Compared to ANOVA, this protocol provided more biological information and was more appropriate for analysis of the data. This approach is therefore recommended in soil seed bank and invasion ecological studies.     

A promising 31P NMR-multivariate analysis approach for the identification of milk phosphorylated metabolites and for rapid authentication of milk samples

A fast and reliable method for the identification of milk from different mammalians was developed by using ³¹P NMR metabolite profile of milk serum coupled to multivariate analysis (PCA and classification models UNEQ, SIMCA and K-NN). Ten milk samples from six different mammalians, relevant to human nutrition (human, cow, donkey, mare, goat, sheep), were analyzed and eight monophosphorylated components were identified and quantified: phosphocreatine (PCr), glycerophosphorylcholine (GPC), glycerophosphorylethanolamine (GPE), N-acetylglucosamine-1-phosphate (NAcGlu-1P), lactose-1-phosphate (Lac-1P), galactose-1-phosphate (Gal-1P), phosphorylcholine (PC), glucose-6-phosphate (Glu-6P). PCA showed interesting clustering based on the animal genus. K-NN can be successfully used to discriminate between donkey and cow samples while UNEQ class-modeling resulted more suitable for compliance verification. Results confirm the natural variability of milk samples among different species. These data highlight the great potentials of NMR/multivariate analysis combined method in the rapid analysis of phosphorylated milk serum metabolites for milk origin assessment and milk adulteration detection.