Study Using Illumina Microarray Technology

Transcription

1 Gene Expression Analysis in Partek Express : An Age Study Using Illumina Microarray Technology This tutorial will provide step-by-step directions for analyzing an Illumina gene expression data set using the Partek Express software package. Topics covered include Creating a new study Editing sample information Importing Illumina data and performing the QC check Visualizing sample-level grouping using PCA Defining differentially expressed genes using ANOVA Description of the Data Set The data set used in this tutorial is based on 51 subjects ran on the Illumina Human Ref-8 BeadChip platform. 26 of the subjects were categorized as Young with an age range of 18 to 28. The other 25 subjects were categorized as Old with an age range of 65 to 84. Skeletal muscle, a type of striated muscle tissue, was obtained via a biopsy from each of the subjects. The skeletal muscle cells had the total RNA extracted, prepped, and ran on the BeadChips producing the data that is used with this tutorial. By running the appropriate statistical analyses on this data set, it is possible to see if there are any transcriptomic changes in the skeletal muscle tissue as people age. The reference for the paper is Melov, S., Tarnopolsky, M.A., Beckman, K., Felkey, K., Hubbard, A. (2007). Resistance Exercise Reverses Aging in Human Skeletal Muscle. PLoS ONE 2(5): e465. doi: /journal.pone It can be found at Importing Illumina Files and Data Data from Illumina s GenomeStudio software package can be exported in a custom Partek report file for seamless importation of your Illumina data into Partek Express. The GenomeStudio Plug-in required for the generation of the custom Partek report file, as well as instructions for its installation, can be found in the Appendix part of this tutorial. Download the data and annotation files for this experiment from For this example, the data files are stored in the C:\Partek Express Demo Data\ Illumina_Age_Data folder Double-click the Partek Express icon to launch the software. Note: if you haven t installed the software, go to the Partek Express update page: to do so Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 1

2 Creating a New Study The creation of a study is the first step for analysis when using Partek Express. This step designates where the data and analysis will be saved on the computer as a.pex file. The.pex file contains all of the information of the study in one file that allows for an easy transfer of the study from one computer to another or when returning to a study later. Select the button to create a new study analyzing Illumina gene expression data Browse to C:\Partek Express Demo Data\ Illumina_Age_Data folder and name the study file as age (Figure 1) Figure 1: Viewing the New Study dialog Select the Illumina_Age_Project.ppj file (Figure 2) Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 2

3 Figure 2: Loading the Partek report file Editing Sample Information The loaded data already contains the sample information (Figure 3). With the sample editor in Partek Express, you can edit sample information, such as adding and deleting attributes, rearranging samples, or ordering sample attributes. Figure 3: Viewing the loaded Illumina Study within Partek Express Confirming Experimental Design Select the column title or any cell in a categorical column to view a histogram showing the distribution of categories in that column Viewing the distribution of the categories can be used to quickly ensure that the correct sample information was assigned to the experiment. A graph of how the samples are distributed across the selected categorical variable is displayed in the bottom pane of Figure 3, which shows 51 Total Samples, 25 Old and 26 Young. Importing Illumina Data and Viewing QC Metrics After editing the sample information, select Next Upon finishing the import, the quality assessment results will be shown in the QC Metrics tab. The results can be viewed in either a Box Plot (Figure 4) or a MA Plot (Figure 5) by selecting the corresponding tab. Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 3

4 Figure 4: Viewing the QC Metrics Box Plot on Log Expression Signal Figure 5: Viewing the QC Metrics MA Plot Viewing the PCA Plot PCA is an excellent method for visualizing high dimensional data by reducing the variation across all of the many thousands of probes being interrogated on the chip into a two or three dimensional representation. In a PCA plot, each point represents a sample (microarray) and corresponds to a row in the Sample Information tab. The positions of the dots are relative to each other. The dots, which are closer to each other, represent samples in which the transcriptome measurements over the whole chip are similar. The dots that are further away from each other represent samples in which the transcriptome measurements Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 4

5 over the whole chip are more dissimilar. Samples that have similar overall gene expression levels will group together into clusters. Identifying separate clusters in a PCA provides valuable information, such as which of the phenotypic variables are driving the major sources of variation within the experiment. One example would be if an experiment only had one factor, treated and untreated. Assuming that all the samples in the data set are the same except for this one factor, it is possible to quickly identify if the treatment had a significant effect on the overall gene expression. If all of the samples clustered together into one group with the two colors mixed equally among the cluster, then there is no distinctive difference between the gene expressions over the samples based upon treatment. However, if the samples cluster into two distinct groups, one cluster containing only treated samples and the other cluster containing only untreated samples, then there is a difference in the gene expression profiles between the treated and untreated samples. If there are multiple factors in an experiment, the factor in which the samples cluster on would likely be the factor with the greatest variation once the ANOVA was run. Without doing any statistical analysis, it is possible to identify the factor having the greatest effect on the overall gene expression of the experiment. Select Next to invoke the PCA plot In the resulting PCA plot, the default color of the dots is dependent on the first column after the filename in the Sample Information tab. In this data set, the first column header is Young/Old, and contains two groups, Old (represented by red dots) and Young (represented by blue dots). Press the mouse middle mouse button/wheel and drag it to rotate the plot or choose the Rotate Mode option ( ). If using the Rotate mode, press and drag the left mouse button to rotate the plot to examine the grouping pattern or outliers of the data on the first 3 principal components (PCs). On PC#3, a loose segregation between the old and young samples is apparent, the old samples lie to the left, and the young samples lie to the right (Figure 6) Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 5

6 Figure 6: Viewing a PCA scatter plot of the Illumina Age data to detect grouping patterns in the data. From this plane, Old lie to the left and Young lie to the right Select the button to reset the rotation to the original position Select the drop down list (Figure 7) then choose 6. BeadChip Figure 7: Coloring by 6. BeadChip The dots are now colored based upon the BeadChip used for each sample (Figure 8). Notice that the samples group together based upon the BeadChip associated with samples. This is an example of a technical batch effect. This technical batch effect can be accounted for by including BeadChip as one of the factors in the ANOVA model allowing the technical noise of the BeadChip to be minimized and allowing for better detection of the biological variation in the experiment. Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 6

7 Figure 8: Viewing a PCA scatter plot of the Illumina Age data colored by BeadChip PCA is an example of exploratory data analysis and is useful for identifying major effects in the data. From this scatter plot exercise, age and BeadChip are shown to have significant sources of variation in the data set based upon the grouping patterns. The variation in Gender is not a significant source of variation as there is no distinct separation of the two groups in the scatter plot. You are left to view this on your own. PCA can also be used to identify sample outliers. If any given sample is not grouping with other replicates, then select the sample in the PCA plot and the row corresponding to the sample will be selected in the Sample Information tab. Right-clicking on that row allows the sample to be deleted from the experiment. In this example, there are no clear anomalous samples. To export a static image of the PCA plot, go to the File menu and select Save Image As, and then select the desired export format, the name, and the location of the resulting exported file. PCA does not provide any specific statistical analysis; in particular, it does not answer the questions regarding which individual genes are being differentially expressed over the factors in the experiment. To discover which genes are differentially expressed, Partek Express will use ANOVA in the next step to provide this information. Identifying Differentially Expressed Genes using ANOVA Analysis of variance (ANOVA) is a very powerful technique for identifying differentially expressed genes and can be used generally in either a simple or a complicated, multi-factor experiment. One of ANOVA s attributes is its applicability to a wide range of use cases. Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 7

8 To set up the ANOVA with Age as the factor, follow these steps: Select Next from the PCA tab; the Detect Differentially Expressed Genes dialog will appear (Figure 9) Drag the effect Age from the Unassigned Effects to the Effect of Interest 1 group Select Next Figure 9: Configuring the Detect Differentially Expressed Genes Dialog The next page of the Detect Differentially Expressed Genes dialog is a summary (Figure 10), which shows what factors have been selected as an effect of interest and what factors are recommended to be included as reduce noise. From the PCA, notice that there is a batch effect on BeadChip. Here, Partek Express also detected that BeadChip can be used to reduce noise. Partek Express automatically assigns batch effects for you based on the correlation and significance test. Cramer s V is used to do the correlation test between the batch and the categorical main factors and uses Pearson correlation coefficient to test the correlation between batch and numeric main factors. Experience shows that the batch probably needs to be excluded from the model if it has a strong correlation with the main effect or interaction since the strong correlation might indicate a confounding or nesting-nested relationship between the batch and the main factor or interaction. The Significance test is used to test whether those batches that passed the correlation test are significant or not. Partek Express uses model selection techniques to pick those batches that improve the model s adjusted R-Squares most. An FDR multiple test correction will be performed and the number of genes that pass the test will be recorded in the Report tab. The percentages of the FDR test are 5% and 10% by default. Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 8

9 Figure 10: Viewing the summary of the Detect Differentially Expressed Genes dialog Select OK to run the ANOVA model Viewing Effect Sizes After the calculation has finished, the Effect Sizes tab will appear. Effect sizes provide information on the importance of each experimental factor to the transcriptome overall. Effect sizes can be displayed as either a bar chart or a pie chart. Configuring the Effect Sizes Bar Chart The effect sizes bar chart provides information on the variation contributed by factors across all test variables in the ANOVA model (Figure 11). The X-axis of the plot represents the factors and the interactions in the ANOVA model; the Y-axis represents the signal to noise ratio. The mean value across the signal to noise ratio of all genes is plotted on the Y-axis. Notice that the Noise bar in the chart is 1. Noise will always be 1 as it is describes the background noise relative to the signal detected in the other factors. Relative to the Noise bar, factors with taller bars represent factors that are more significant. Bars at or near Noise represent factors that are not as significant to the transcriptome overall. On average across all genes, BeadChip is the biggest source of variation in this data set. To export the effect sizes image, select Save Image As from the File menu. Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 9

10 Figure 11: Viewing the Effect Sizes Bar Configuring the Effect Sizes Pie Chart The Effect Sizes pie chart shows a comparison of importance between different factor effects (Figure 12). Each section of the pie chart is labeled with the name of the factor and a percentage of the pie contained in the corresponding slice. Larger pieces of the pie indicate more significant factors, while factors at or near the size of the Noise slice are not significant to the transcriptome overall. Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 10

11 Figure 12: Viewing the Effect Sizes Pie Chart Viewing and Interpreting the Gene Significance Spreadsheet Select Next, and the Gene Significance table will appear showing individual gene results (Figure 13) One of the most critical pieces of information contained in the Gene Significance table is the p-value per gene per categorical variable. A p-value is a test statistic (between zero and one) used to rank significance of results of starting with the null hypothesis that a gene is similarly expressed across conditions, which means the smaller the p-value for a given gene, the more likely the gene shows differential expression across the given categorical variables. A dot plot is shown in the right pane of this tab for the currently selected row. In the dot plot, each dot is an individual sample data point. The X-Axis represents the different types and the Y-Axis displays the log2 expression level of the gene. The box & whiskers are grouped by Young/Old and colored by Young/Old. Selecting different options in the dropdown menus ( ) will change the settings. The data is converted into log2 space to ensure that the data is more normally distributed so that an ANOVA test can be performed. When using statistical tests such as ANOVA or t-tests, one of the assumptions of the test is that the data is normalized ; with the log2 transformation, this assumption is met. When data is in log2 space, it is important to remember that the scale is typically between zero and 16, and that any increment change of one represents a two-fold change in abundance. For instance, if a gene changes from 6 in Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 11

12 one condition to 8 in another condition a four-fold change between the two conditions has occurred. The first p-value column in the Gene Significance table is Age. This column should be automatically sorted ascending with the genes having the smallest p-values at the top. Genes in the top rows are the most significant differentially expressed genes across the variables in Age in the experiment. In this example, there are only two classes within Type Old and Young. Focusing on the top gene, USP54, a detailed view of the individual signal intensities for this gene can be viewed in the dot plot in the left pane. The median of USP54 in the Old samples is around 10.3, while the median of Young samples is around 9.0. Figure 13: Viewing the Gene Significance Spreadsheet and Dot Plot, grouped by Type and colored by Tissue A search bar is provided for gene name searches at the top of the Gene Significance Estimates tab. To search the spreadsheet, type the search string, such as gene symbol or gene title, in the entry box and select Next or Previous. You can also specify if you want the search to be case sensitive and/or to match the whole cell by checking the respective check boxes. Any column of the spreadsheet can be sorted ascending or descending by leftclicking on the column header. This is useful in searching for the lowest or highest values in a column or to sort a text column alphabetically. Power Analysis and Pathway Analysis This tutorial focuses the statistical analysis on the continuous variable Age. Fold-changes cannot be generated with a continuous variable. Since both Power Analysis and Pathway Analysis need the fold-changes as the input, this tutorial will end here without these 2 optional analyses. For more information about Power Analysis and Pathway Analysis, refer to the Analyzing Disease vs. Normal in Partek Express : A Down Syndrome Study tutorial located at Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 12

13 Appendix: Using Partek s Report Plug-in for Illumina s GenomeStudio to Export Gene Expression Data This appendix describes how to export gene expression data using Partek s Report Plug-in for the Illumina GenomeStudio Gene Expression Module for use in Partek Express. By using the GenomeStudio plug-in, you can export your data into a project that Partek can open directly. It is the fastest and most consistent way to get fully annotated Illumina gene expression data into Partek. Installing the Plug-in Open Illumina GenomeStudio, and download the plug-in from the Portal Extract the PartekReport.GeneExpression.dll file from the distributed zip file Move the DLL file into GenomeStudio s Gene Expression Module s report plug-in folder (Create the folder if it does not exist) C:/ProgramFiles/Illumina/GenomeStudio/Modules/BSGX/ReportPlugins/Partek Report (Figure 1a) Note: If you have old versions of the plug-in, please remove them from the GenomeStudio folder hierarchy Figure 1a: Placing the Partek Report DLL in the appropriate directory Exporting from GenomeStudio Select Analysis > Reports from the main menu Select Custom Report Choose Partek Report Plug-in from the list of custom plug-ins (Figure 2a) Configure export options based on information below Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 13

14 Figure 2a: Configuring the GenomeStudio Gene Expression Report dialog Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 14

15 Export Options Annotation AnnotationName LaunchPartek SampleInformation Type Table 1: Export options The level of detail you would like your exported annotation to have. Minimal will contain the required information to do data analysis, and Full will contain everything from GenomeStudio. This should be equivalent to the name of the bgx file you imported the data with. Common values are HumanRef8v2 or HumanWG-6v2. If you don t know this information, you should just use something unique to your data set. After the plug-in has completed exporting, you can launch Partek Genomics Suite software. If you only installed Partek Express software, select false. The level of detail you would like your exported sample information to have. If you have entered data from your experiment into the samples table, you should select Full. Illumina s gene expression platforms output multiple probe intensities for each gene. The probe level output these intensities directly. Gene level output averages all the probe intensities for each gene. It is strongly recommended to put the exported files in their own folder. This allows you to move the folder instead of all the files individually Select OK to start the export process The result of the export process is a project file (.ppj), which you can open in Partek Genomics Suite or in Partek Express. The project file will open the signal intensities data and associate the annotation information to the intensities file. All intensities are log transformed. Annotation information describes the probe or gene markers, which you have just created data for. The plug-in also creates a manifest file describing the files and date of creation. Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 15

16 Figure 3a: Viewing the exported files from the GenomeStudio project Copyright 2010 by Partek Incorporated. All Rights Reserved. Reproduction of this material without express written consent from Partek Incorporated is strictly prohibited. Gene Expression Analysis in Partek Express: An Age Study Using Illumina Microarray Technology 16