Classifying aneuploidy in genotype intensity data using deep learning.

Aneuploidy is the loss or gain of one or more chromosomes. Although it is a rare phenomenon in liveborn individuals, it is observed in livestock breeding populations. These breeding populations are often routinely genotyped and the genotype intensity data from single nucleotide polymorphism (SNP) arrays can be exploited to identify aneuploidy cases. This identification is a time-consuming and costly task, because it is often performed by visual inspection of the data per chromosome, usually done in plots of the intensity data by an expert. Therefore, we wanted to explore the feasibility of automated image classification to replace (part of) the visual detection procedure for any diploid species. The aim of this study was to develop a deep learning Convolutional Neural Network (CNN) classification model based on chromosome level plots of SNP array intensity data that can classify the images into disomic, monosomic and trisomic cases. A multispecies dataset enriched for aneuploidy cases was collected containing genotype intensity data of 3321 disomic, 1759 monosomic and 164 trisomic chromosomes. The final CNN model had an accuracy of 99.9%, overall precision was 1, recall was 0.98 and the F1 score was 0.99 for classifying images from intensity data. The high precision assures that cases detected are most likely true cases, however, some trisomy cases may be missed (the recall of the class trisomic was 0.94). This supervised CNN model performed much better than an unsupervised k-means clustering, which reached an accuracy of 0.73 and had especially difficult to classify trisomic cases correctly. The developed CNN classification model provides high accuracy to classify aneuploidy cases based on images of plotted X and Y genotype intensity values. The classification model can be used as a tool for routine screening in large diploid populations that are genotyped to get a better understanding of the incidence and inheritance, and in addition, avoid anomalies in breeding candidates.

Plasticity of intestinal gene expression profile signatures reflected by nutritional interventions in piglets.

BACKGROUND: Immediately after birth, the porcine intestine rapidly develops morphologically, functionally, and immunologically. The jejunum, the second part of the small intestine, is of importance for nutrient uptake and immune surveillance. To study the early postnatal development of the jejunum, a meta-analysis was performed on different transcriptomic datasets. These datasets were acquired from different experimental in-house studies or from experiments described in literature of porcine jejunum mucosa. Gene expression was measured under different experimental interventions, such as nutritional intervention, at various time-points (age). RESULTS: The studies included in the meta-analysis provided gene expression data for various time-points (piglet ages) for piglets that had received a treatment versus control piglets. In separate studies, treatments were administered to the sow (i.e. amoxicillin), or nutritional supplementation directly to the piglets with medium chain fatty acids (MCFAs), and oral administration of fructooligosaccharides (FOS) or a high dose of zinc-oxide, respectively. In the meta-analysis, genes were grouped into 16 clusters according to their temporal gene expression profiles for control piglets, i.e. the changes of gene expression level over time. Functional analysis showed that these temporal profile clusters had different dominant processes, such as immune related processes or barrier function. Transcriptomics data of treatment piglets was subsequently superimposed over the control temporal profiles. In this way we could investigate which temporal profile clusters (and which biological processes) were modulated by the treatments. Interestingly, not all 16 temporal profiles were modulated. CONCLUSIONS: We showed that it is possible to re-use (publicly available) transcriptomics data and produce temporal gene expression profiles for control piglets with overexpression of genes representing specific biological processes. Subsequently, by superimposing gene expression data from (nutritional) intervention studies we observed deviations from some of these reference profile(s) and thus the plasticity of the system. By employing this meta-analysis approach we highlighted the importance of birth and weaning and the underlying biological processes.

Impact of merging commercial breeding lines on the genetic diversity of Landrace pigs.

BACKGROUND: The pig breeding industry has undergone a large number of mergers in the past decades. Various commercial lines were merged or discontinued, which is expected to reduce the genetic diversity of the pig species. The objective of the current study was to investigate the genetic diversity of different former Dutch Landrace breeding lines and quantify their relationship with the current Dutch Landrace breed that originated from these lines. RESULTS: Principal component analysis clearly divided the former Landrace lines into two main clusters, which are represented by Norwegian/Finnish Landrace lines and Dutch Landrace lines. Structure analysis revealed that each of the lines that are present in the Dutch Gene bank has a unique genetic identity. The current Dutch Landrace breed shows a high level of admixture and is closely related to the six former lines. The Dumeco N-line, which is conserved in the Dutch Gene bank, is poorly represented in the current Dutch Landrace. All seven lines (the six former and the current line) contribute almost equally to the genetic diversity of the Dutch Landrace breed. As expected, the current Dutch Landrace breed comprises only a small proportion of unique genetic diversity that was not present in the other lines. The genetic diversity level, as measured by Eding's core set method, was equal to 0.89 for the current Dutch Landrace breed, whereas total genetic diversity across the seven lines, measured by the same method, was equal to 0.99. CONCLUSIONS: The current Dutch Landrace breed shows a high level of admixture and is closely related to the six former Dutch Landrace lines. Merging of commercial Landrace lines has reduced the genetic diversity of the Landrace population in the Netherlands, although a large proportion of the original variation is maintained. Thus, our recommendation is to conserve breeding lines in a gene bank before they are merged.

The impact of using old germplasm on genetic merit and diversity-A cattle breed case study.

Artificial selection and high genetic gains in livestock breeds led to a loss of genetic diversity. Current genetic diversity conservation actions focus on long-term maintenance of breeds under selection. Gene banks play a role in such actions by storing genetic materials for future use and the recent development of genomic information is facilitating characterization of gene bank material for better use. Using the Meuse-Rhine-Issel Dutch cattle breed as a case study, we inferred the potential role of germplasm of old individuals for genetic diversity conservation of the current population. First, we described the evolution of genetic merit and diversity over time and then we applied the optimal contribution (OC) strategy to select individuals for maximizing genetic diversity, or maximizing genetic merit while constraining loss of genetic diversity. In the past decades, genetic merit increased while genetic diversity decreased. Genetic merit and diversity were both higher in an OC scenario restricting the rate of inbreeding when old individuals were considered for selection, compared to considering only animals from the current population. Thus, our study shows that gene bank material, in the form of old individuals, has the potential to support long-term maintenance and selection of breeds.

Study of potential health effects of electromagnetic fields of telephony and Wi-Fi, using chicken embryo development as animal model.

The objective of this study is to investigate possible biological effects of radiofrequency electromagnetic fields (RF-EMF) as used in modern wireless telecommunication in a well-controlled experimental environment using chicken embryo development as animal model. Chicken eggs were incubated under continuous experimental exposure to GSM (1.8 GHz), DECT (1.88 GHz), UMTS (2.1 GHz), and WLAN (5.6 GHz) radiation, with the appropriate modulation protocol, using a homogeneous field distribution at a field strength of approximately 3 V/m, representing the maximum field level in a normal living environment. Radiation-shielded exposure units/egg incubators were operating in parallel for exposed and control eggs in a climatized homogeneous environment, using 450 eggs per treatment in three successive rounds per treatment. Dosimetry of the exposure (field characteristics and specific absorption rate) were studied. Biological parameters studied included embryo death during incubation, hatching percentage, and various morphological and histological parameters of embryos and chicks and their organs, and gene expression profiles of embryos on day 7 and day 18 of incubation by microarray and qPCR. No conclusive evidence was found for induced embryonic mortality or malformations by exposure to the used EMFs, or for effects on the other measured parameters. Estimated differences between treatment groups were always small and the effect of treatment was not significant. In a statistical model that ignored possible interaction between rounds and exposure units, some of the many pairwise comparisons of exposed versus control had P-values lower than 0.05, but were not significant after correction for multiple testing. Bioelectromagnetics. 38:186-203, 2017. © 2017 Wiley Periodicals, Inc.

Accuracy of imputation to whole-genome sequence data in Holstein Friesian cattle.

BACKGROUND: The use of whole-genome sequence data can lead to higher accuracy in genome-wide association studies and genomic predictions. However, to benefit from whole-genome sequence data, a large dataset of sequenced individuals is needed. Imputation from SNP panels, such as the Illumina BovineSNP50 BeadChip and Illumina BovineHD BeadChip, to whole-genome sequence data is an attractive and less expensive approach to obtain whole-genome sequence genotypes for a large number of individuals than sequencing all individuals. Our objective was to investigate accuracy of imputation from lower density SNP panels to whole-genome sequence data in a typical dataset for cattle. METHODS: Whole-genome sequence data of chromosome 1 (1737 471 SNPs) for 114 Holstein Friesian bulls were used. Beagle software was used for imputation from the BovineSNP50 (3132 SNPs) and BovineHD (40 492 SNPs) beadchips. Accuracy was calculated as the correlation between observed and imputed genotypes and assessed by five-fold cross-validation. Three scenarios S40, S60 and S80 with respectively 40%, 60%, and 80% of the individuals as reference individuals were investigated. RESULTS: Mean accuracies of imputation per SNP from the BovineHD panel to sequence data and from the BovineSNP50 panel to sequence data for scenarios S40 and S80 ranged from 0.77 to 0.83 and from 0.37 to 0.46, respectively. Stepwise imputation from the BovineSNP50 to BovineHD panel and then to sequence data for scenario S40 improved accuracy per SNP to 0.65 but it varied considerably between SNPs. CONCLUSIONS: Accuracy of imputation to whole-genome sequence data was generally high for imputation from the BovineHD beadchip, but was low from the BovineSNP50 beadchip. Stepwise imputation from the BovineSNP50 to the BovineHD beadchip and then to sequence data substantially improved accuracy of imputation. SNPs with a low minor allele frequency were more difficult to impute correctly and the reliability of imputation varied more. Linkage disequilibrium between an imputed SNP and the SNP on the lower density panel, minor allele frequency of the imputed SNP and size of the reference group affected imputation reliability.

Prioritization of candidate genes for cattle reproductive traits, based on protein-protein interactions, gene expression, and text-mining.

Reproduction is of significant economic importance in dairy cattle. Improved understanding of mechanisms that control estrous behavior and other reproduction traits could help in developing strategies to improve and/or monitor these traits. The objective of this study was to predict and rank genes and processes in brain areas and pituitary involved in reproductive traits in cattle using information derived from three different data sources: gene expression, protein-protein interactions, and literature. We identified 59, 89, 53, 23, and 71 genes in bovine amygdala, dorsal hypothalamus, hippocampus, pituitary, and ventral hypothalamus, respectively, potentially involved in processes underlying estrus and estrous behavior. Functional annotation of the candidate genes points to a number of tissue-specific processes of which the "neurotransmitter/ion channel/synapse" process in the amygdala, "steroid hormone receptor activity/ion binding" in the pituitary, "extracellular region" in the ventral hypothalamus, and "positive regulation of transcription/metabolic process" in the dorsal hypothalamus are most prominent. The regulation of the functional processes in the various tissues operate at different biological levels, including transcriptional, posttranscriptional, extracellular, and intercellular signaling levels.

Meta-analysis of chicken--salmonella infection experiments.

BACKGROUND: Chicken meat and eggs can be a source of human zoonotic pathogens, especially Salmonella species. These food items contain a potential hazard for humans. Chickens lines differ in susceptibility for Salmonella and can harbor Salmonella pathogens without showing clinical signs of illness. Many investigations including genomic studies have examined the mechanisms how chickens react to infection. Apart from the innate immune response, many physiological mechanisms and pathways are reported to be involved in the chicken host response to Salmonella infection. The objective of this study was to perform a meta-analysis of diverse experiments to identify general and host specific mechanisms to the Salmonella challenge. RESULTS: Diverse chicken lines differing in susceptibility to Salmonella infection were challenged with different Salmonella serovars at several time points. Various tissues were sampled at different time points post-infection, and resulting host transcriptional differences investigated using different microarray platforms. The meta-analysis was performed with the R-package metaMA to create lists of differentially regulated genes. These gene lists showed many similarities for different chicken breeds and tissues, and also for different Salmonella serovars measured at different times post infection. Functional biological analysis of these differentially expressed gene lists revealed several common mechanisms for the chicken host response to Salmonella infection. The meta-analysis-specific genes (i.e. genes found differentially expressed only in the meta-analysis) confirmed and expanded the biological functional mechanisms. CONCLUSIONS: The meta-analysis combination of heterogeneous expression profiling data provided useful insights into the common metabolic pathways and functions of different chicken lines infected with different Salmonella serovars.

Selection and Drift: A Comparison between Historic and Recent Dutch Friesian Cattle and Recent Holstein Friesian Using WGS Data.

Over the last century, genetic diversity in many cattle breeds has been affected by the replacement of traditional local breeds with just a few milk-producing breeds. In the Netherlands, the local Dutch Friesian breed (DF) has gradually been replaced by the Holstein Friesian breed (HF). The objective of this study is to investigate genomewide genetic diversity between a group of historically and recently used DF bulls and a group of recently used HF bulls. Genetic material of 12 historic (hDF), 12 recent DF bulls (rDF), and 12 recent HF bulls (rHF) in the Netherlands was sequenced. Based on the genomic information, different parameters-e.g., allele frequencies, inbreeding coefficient, and runs of homozygosity (ROH)-were calculated. Our findings showed that a large amount of diversity is shared between the three groups, but each of them has a unique genetic identity (12% of the single nucleotide polymorphisms were group-specific). The rDF is slightly more diverged from rHF than hDF. The inbreeding coefficient based on runs of homozygosity (Froh) was higher for rDF (0.24) than for hDF (0.17) or rHF (0.13). Our results also displayed the presence of several genomic regions that differentiated between the groups. In addition, thirteen, forty-five, and six ROH islands were identified in hDF, rDF, and rHF, respectively. The genetic diversity of the DF breed reduced over time, but this did not lead to higher inbreeding levels-especially, inbreeding due to recent ancestors was not increased.

Retriever and Pointer: Software to Evaluate Inbreeding and Genetic Management in Captive Populations.

The Retriever and Pointer software has been developed for genetic management of (small) captive populations The Retriever program uses as input pedigree data and extracts data on population structure that determine inbreeding rates such as skewness of sire contributions. Levels and rates of inbreeding and kinship and effective population sizes are determined as well. Data on population structure can be used as input for the Pointer program. This program uses stochastic simulation to evaluate a population and provides expected levels and rates of inbreeding and kinship, and optionally allelic diversity. The user can simulate different options for genetic management such as sire restrictions, restrictions on inbreeding levels, mean kinships and breeding circles. Both Retriever and Pointer can analyze populations with subpopulations and different rates of exchange between them. Although originally devised for dogs, the software can be, and has been, used for any captive population including livestock and zoo populations, and a number of examples are provide The pointer software is also suitable in education where students may generate their own populations and evaluate effects of different population structures and genetic management on genetic diversity. Input is provided via a graphical user interface. The software can be downloaded for free.

Boosted trees to predict pneumonia, growth, and meat percentage of growing-finishing pigs1.

In pig production, efficiency is benefiting from uniform growth in pens resulting in single deliveries from a pen of possibly all animals in the targeted weight range. Abnormalities, like pneumonia or aberrant growth, reduce production efficiency as it reduces the uniformity and might cause multiple deliveries per batch and pigs delivered with a low meat yield or outside the targeted weight range. Early identification of pigs prone to develop these abnormalities, for example, at the onset of the growing-finishing phase, would help to prevent heterogeneous pens through management interventions. Data about previous production cycles at the farm combined with data from the piglet's own history may help in identifying these abnormalities. The aim of this study, therefore, was to predict at the onset of the growing-finishing phase, that is, at 3 mo in advance, deviant pigs at slaughter with a machine-learning technique called boosted trees. The dataset used was extracted from the farm management system of a research center. It contained over 70,000 records of individual pigs born between 2004 and 2016, including information on, for example, offspring, litter size, transfer dates between production stages, their respective locations within the barns, and individual live-weights at several production stages. Results obtained on an independent test set showed that at a 90% specificity rate, the sensitivity was 16% for low meat percentage, 20% for pneumonia and 36% for low lifetime growth rate. For low lifetime growth rate, this meant an almost three times increase in positive predictive value compared to the current situation. From these results, it was concluded that routine performance information available at the onset of the growing-finishing phase combined with data about previous production cycles formed a moderate base to identify pigs prone to develop pneumonia (AUC > 0.60) and a good base to identify pigs prone to develop growth aberrations (AUC > 0.70) during the growing-finishing phase. The mentioned information, however, was not a sufficient base to identify pigs prone to develop low meat percentage (AUC < 0.60). The shown ability to identify growth aberrations and pneumonia can be considered a good first step towards the development of an early warning system for pigs in the growing-finishing phase.

Assessment of sire contribution and breed-of-origin of alleles in a three-way crossbred broiler dataset.

Broiler breeding programs rely on crossbreeding. With genomic selection, widespread use of crossbred performance in breeding programs comes within reach. Commercial crossbreds, however, may have unknown pedigrees and their genomes may include DNA from 2 to 4 different breeds. Our aim was, for a broiler dataset with a limited number of sires having both purebred and crossbred offspring generated using natural mating, to rapidly derive parentage, assess the distribution of the sire contribution to the offspring generation, and to assess breed-of-origin of alleles in crossbreds. The dataset contained genotypes for 56,075 SNPs for 5,882 purebred and 10,943 3-way crossbred offspring generated by natural mating of 164 purebred sires to 1,016 purebred and 1,386 F1 crossbred hens. Using our algorithm FindParents, joint parentage derivation for the offspring and parent generations required only 1 m 29 s to retrieve parentage for 20,253 animals considering 4,504 possible parents. FindParents was similarly accurate as a maximum likelihood based method, apart from situations where settings of FindParents did not match the genotyping error rate in the data. Numbers of offspring per sire had a very skewed distribution, ranging from 1 to 270 crossbreds and 1 to 154 purebreds. Derivation of breed-of-origin of alleles relied on phasing all genotypes, including 8,205, 372, and 720 animals from the 3 pure lines involved, and allocating haplotypes in the crossbreds to purebred lines based on observed frequencies in the purebred lines. Breed-of-origin could be derived for 96.94% of the alleles of the 1,386 F1 crossbred hens and for 91.88% of the alleles of the 10,943 3-way crossbred offspring, of which 49.49% to the sire line. The achieved percentage of assignment to the sire line was sufficient to proceed with subsequent analyses requiring only the breed-of-origin of the paternal alleles to be known. Although required number of animals may be population dependent, to increase the total percentage of assigned alleles, it seems advisable to use at least approx. 1,000 genotyped purebred animals for each of the lines involved.

Biochemical pathways analysis of microarray results: regulation of myogenesis in pigs.

BACKGROUND: Combining microarray results and biological pathway information will add insight into biological processes. Pathway information is widely available in databases through the internet. Mammalian muscle formation has been previously studied using microarray technology in pigs because these animals are an interesting animal model for muscle formation due to selection for increased muscle mass. Results indicated regulation of the expression of genes involved in proliferation and differentiation of myoblasts, and energy metabolism. The aim of the present study was to analyse microarrays studying myogenesis in pigs. It was necessary to develop methods to search biochemical pathways databases. RESULTS: PERL scripts were developed that used the names of the genes on the microarray to search databases. Synonyms of gene names were added to the list by searching the Gene Ontology database. The KEGG database was searched for pathway information using this updated gene list. The KEGG database returned 88 pathways. Most genes were found in a single pathway, but others were found in up to seven pathways. Combining the pathways and the microarray information 21 pathways showed sufficient information content for further analysis. These pathways were related to regulation of several steps in myogenesis and energy metabolism. Pathways regulating myoblast proliferation and muscle fibre formation were described. Furthermore, two networks of pathways describing the formation of the myoblast cytoskeleton and regulation of the energy metabolism during myogenesis were presented. CONCLUSION: Combining microarray results and pathways information available through the internet provide biological insight in how the process of porcine myogenesis is regulated.

Whole-genome sequencing of 234 bulls facilitates mapping of monogenic and complex traits in cattle.

The 1000 bull genomes project supports the goal of accelerating the rates of genetic gain in domestic cattle while at the same time considering animal health and welfare by providing the annotated sequence variants and genotypes of key ancestor bulls. In the first phase of the 1000 bull genomes project, we sequenced the whole genomes of 234 cattle to an average of 8.3-fold coverage. This sequencing includes data for 129 individuals from the global Holstein-Friesian population, 43 individuals from the Fleckvieh breed and 15 individuals from the Jersey breed. We identified a total of 28.3 million variants, with an average of 1.44 heterozygous sites per kilobase for each individual. We demonstrate the use of this database in identifying a recessive mutation underlying embryonic death and a dominant mutation underlying lethal chrondrodysplasia. We also performed genome-wide association studies for milk production and curly coat, using imputed sequence variants, and identified variants associated with these traits in cattle.

Globaltest and GOEAST: two different approaches for Gene Ontology analysis.

BACKGROUND: Gene set analysis is a commonly used method for analysing microarray data by considering groups of functionally related genes instead of individual genes. Here we present the use of two gene set analysis approaches: Globaltest and GOEAST.Globaltest is a method for testing whether sets of genes are significantly associated with a variable of interest. GOEAST is a freely accessible web-based tool to test GO term enrichment within given gene sets. The two approaches were applied in the analysis of gene lists obtained from three different contrasts in a microarray experiment conducted to study the host reactions in broilers following Eimeria infection. RESULTS: The Globaltest identified significantly associated gene sets in one of the three contrasts made in the microarray experiment whereas the functional analysis of the differentially expressed genes using GOEAST revealed enriched GO terms in all three contrasts. CONCLUSION: Globaltest and GOEAST gave different results, probably due to the different algorithms and the different criteria used for evaluating the significance of GO terms.

Methods for interpreting lists of affected genes obtained in a DNA microarray experiment.

BACKGROUND: The aim of this paper was to describe and compare the methods used and the results obtained by the participants in a joint EADGENE (European Animal Disease Genomic Network of Excellence) and SABRE (Cutting Edge Genomics for Sustainable Animal Breeding) workshop focusing on post analysis of microarray data. The participating groups were provided with identical lists of microarray probes, including test statistics for three different contrasts, and the normalised log-ratios for each array, to be used as the starting point for interpreting the affected probes. The data originated from a microarray experiment conducted to study the host reactions in broilers occurring shortly after a secondary challenge with either a homologous or heterologous species of Eimeria. RESULTS: Several conceptually different analytical approaches, using both commercial and public available software, were applied by the participating groups. The following tools were used: Ingenuity Pathway Analysis, MAPPFinder, LIMMA, GOstats, GOEAST, GOTM, Globaltest, TopGO, ArrayUnlock, Pathway Studio, GIST and AnnotationDbi. The main focus of the approaches was to utilise the relation between probes/genes and their gene ontology and pathways to interpret the affected probes/genes. The lack of a well-annotated chicken genome did though limit the possibilities to fully explore the tools. The main results from these analyses showed that the biological interpretation is highly dependent on the statistical method used but that some common biological conclusions could be reached. CONCLUSION: It is highly recommended to test different analytical methods on the same data set and compare the results to obtain a reliable biological interpretation of the affected genes in a DNA microarray experiment.

Analysis of a simulated microarray dataset: comparison of methods for data normalisation and detection of differential expression (open access publication).

Microarrays allow researchers to measure the expression of thousands of genes in a single experiment. Before statistical comparisons can be made, the data must be assessed for quality and normalisation procedures must be applied, of which many have been proposed. Methods of comparing the normalised data are also abundant, and no clear consensus has yet been reached. The purpose of this paper was to compare those methods used by the EADGENE network on a very noisy simulated data set. With the a priori knowledge of which genes are differentially expressed, it is possible to compare the success of each approach quantitatively. Use of an intensity-dependent normalisation procedure was common, as was correction for multiple testing. Most variety in performance resulted from differing approaches to data quality and the use of different statistical tests. Very few of the methods used any kind of background correction. A number of approaches achieved a success rate of 95% or above, with relatively small numbers of false positives and negatives. Applying stringent spot selection criteria and elimination of data did not improve the false positive rate and greatly increased the false negative rate. However, most approaches performed well, and it is encouraging that widely available techniques can achieve such good results on a very noisy data set.

The EADGENE Microarray Data Analysis Workshop (open access publication).

Microarray analyses have become an important tool in animal genomics. While their use is becoming widespread, there is still a lot of ongoing research regarding the analysis of microarray data. In the context of a European Network of Excellence, 31 researchers representing 14 research groups from 10 countries performed and discussed the statistical analyses of real and simulated 2-colour microarray data that were distributed among participants. The real data consisted of 48 microarrays from a disease challenge experiment in dairy cattle, while the simulated data consisted of 10 microarrays from a direct comparison of two treatments (dye-balanced). While there was broader agreement with regards to methods of microarray normalisation and significance testing, there were major differences with regards to quality control. The quality control approaches varied from none, through using statistical weights, to omitting a large number of spots or omitting entire slides. Surprisingly, these very different approaches gave quite similar results when applied to the simulated data, although not all participating groups analysed both real and simulated data. The workshop was very successful in facilitating interaction between scientists with a diverse background but a common interest in microarray analyses.

Analysis of the real EADGENE data set: multivariate approaches and post analysis (open access publication).

The aim of this paper was to describe, and when possible compare, the multivariate methods used by the participants in the EADGENE WP1.4 workshop. The first approach was for class discovery and class prediction using evidence from the data at hand. Several teams used hierarchical clustering (HC) or principal component analysis (PCA) to identify groups of differentially expressed genes with a similar expression pattern over time points and infective agent (E. coli or S. aureus). The main result from these analyses was that HC and PCA were able to separate tissue samples taken at 24 h following E. coli infection from the other samples. The second approach identified groups of differentially co-expressed genes, by identifying clusters of genes highly correlated when animals were infected with E. coli but not correlated more than expected by chance when the infective pathogen was S. aureus. The third approach looked at differential expression of predefined gene sets. Gene sets were defined based on information retrieved from biological databases such as Gene Ontology. Based on these annotation sources the teams used either the GlobalTest or the Fisher exact test to identify differentially expressed gene sets. The main result from these analyses was that gene sets involved in immune defence responses were differentially expressed.

Analysis of the real EADGENE data set: comparison of methods and guidelines for data normalisation and selection of differentially expressed genes (open access publication).

A large variety of methods has been proposed in the literature for microarray data analysis. The aim of this paper was to present techniques used by the EADGENE (European Animal Disease Genomics Network of Excellence) WP1.4 participants for data quality control, normalisation and statistical methods for the detection of differentially expressed genes in order to provide some more general data analysis guidelines. All the workshop participants were given a real data set obtained in an EADGENE funded microarray study looking at the gene expression changes following artificial infection with two different mastitis causing bacteria: Escherichia coli and Staphylococcus aureus. It was reassuring to see that most of the teams found the same main biological results. In fact, most of the differentially expressed genes were found for infection by E. coli between uninfected and 24 h challenged udder quarters. Very little transcriptional variation was observed for the bacteria S. aureus. Lists of differentially expressed genes found by the different research teams were, however, quite dependent on the method used, especially concerning the data quality control step. These analyses also emphasised a biological problem of cross-talk between infected and uninfected quarters which will have to be dealt with for further microarray studies.