Developed by BioDiscovery, Inc. DualChip evaluation software User Manual Version 1.1

Transcription

1 Developed by BioDiscovery, Inc. DualChip evaluation software User Manual Version 1.1

2 1

3 Table of contents 1. INTRODUCTION SCOPE OF DELIVERY INSTALLATION PROCEDURES SYSTEM REQUIREMENTS DUALCHIP EVALUATION SOFTWARE INSTALLATION COMPATIBILITY OF DUALCHIP EVALUATION SOFTWARE WITH QUANTIFICATION SOFTWARE DATA ANALYSIS WORKFLOW INTRODUCTION MICROARRAY CONTROL PROBES RAW DATA TRANSFORMATION DATA SELECTION AND REPLICATES DATA MERGING DATA VALIDATION NORMALIZATION Normalization based on Internal Standard Normalization based on House keeping genes Example STATISTICAL SIGNIFICANCE OF GENE EXPRESSION RATIOS MULTIPLE SCANNING DATA WORK PROCEDURE STEP 1: MODE 1: DUALCHIP EVALUATION SOFTWARE LAUNCHING STEP 2: MODE 1: QUANTIFICATION DATA FOR REFERENCE AND EXPERIMENT LOADING How to load data from supported quantification software How to load data from unsupported quantification software: generic text file loader Load GeneID file STEP 3: MODE 1: DATA ANALYSIS LAUNCHING STEP 4: MODE 1: INTERNAL STANDARDS NORMALIZATION Internal standards table Color code Selection Further interaction STEP 5: MODE 1: HOUSE KEEPING GENES NORMALIZATION House keeping gene table Color code Selection Further interaction STEP 6: MODE 1: GENE EXPRESSION PROFILE TABLES: REPORT OF THE ANALYSIS Ratio tables Color code Normalization data and negative hybridization controls mean Further interaction STEP 7: MODE 1: SCATTER PLOTS AND HISTOGRAMS LOAD SAVED REPORT SETTINGS - PARAMETERS STEP 8: MODE 2: ANALYSIS OF EXPERIMENT REPLICATES Automatic selection Color code Selection change Further interaction COLOR CODE SUMMARY

4 1. Introduction This guide presents the data analysis workflow implemented in the DualChip evaluation software and the various steps carried out when performing this analysis to obtain gene expression results. DualChip evaluation software provides an easy-to-use, streamlined gene expression analysis tool for DualChip microarrays. The software enables the user to receive a gene expression profile for comparison of an experiment and its related reference. The workflow of this analysis, specially designed for DualChip microarrays experiments, includes capabilities for the extraction of gene expression intensities from replicate quantification data, the normalization of these raw data, the identification of significantly expressed genes of an experiment in comparison to a reference, the visualization and the saving of the results. Semi-quantitative analysis can only be performed if the fluorescence intensities of reference and experiment are within an acceptable range corresponding to the linear part of the scanning curve. This means that only genes that are not saturated or are very low (lower than 2 times the local background) can be used for quantitative analysis. To increase the number of genes with quantitatively significant ratios, we recommend scanning the DualChip with different scanner settings (PMT settings for scanners equipped with a photomultiplier tube). This increases the dynamic range of the system. An intermediate scanner setting is first chosen to analyze medium expressed genes. Highly expressed genes can be accurately scanned at a low scanner setting and low expressed genes at a high scanner setting. The DualChip evaluation software procedure allows simultaneous analysis of arrays scanned at different scanner settings. Additionally, the software is designed for the comparison of several replicates of gene expression profiles obtained by the data analysis workflow. 3

5 The following diagram describes classical analysis of gene expression experiments: Step 1: Scan reference and experiment arrays with different scanner settings Step 2: Quantification Quantify the images obtained in step 1 Scanning and quantification software Step 3: Load quantification data for reference and experiment Step 4: Normalize reference and experiment data with Internal Standards and house keeping genes Repeat for each experiment/ replicate DualChip evaluation software Mode 1 Step 5: Save gene expression results Step 6: Analyse experiment replicates DualChip evaluation software Mode 2 2. Delivery package The delivery package includes the following: - 1 CD-ROM (DualChip evaluation software) including user manual - 1 quickstart guide - 1 user manual 4

6 3. Installation procedure 3.1. System requirements - PC with Pentium processor, min. 266 MHz, recommended 700 MHz + - RAM: minimum 256 MB installed, 512 MB recommended - Video: minimum SVGA (800x600), recommended XVGA (1280x1024) - Internet connection recommended (link to gene banks) - Hard disk: 200 MB free disk space for installation, more for image storage - Operating system: Microsoft Windows 98, 98 Second Edition, Millennium, 2000, or XP Important note: user MUST BE logged in as administrator of the computer for software installation DualChip evaluation software installation To install the software: Insert DualChip evaluation software CD into the CD-ROM drive. The installation screen for DualChip evaluation software should be displayed automatically. If the installation screen is not displayed automatically, start the program using: Start>My Computer. Now double-click the CD-ROM drive icon and the program to be executed. Review the version information and click Continue to unpack the installation files. The name of your program folder is automatically: c:\dualchip Evaluation Software\. This will be the location of the DualChip evaluation software shortcuts in your Start menu. Click the "Next button. Click the "Yes" button to accept the License Agreement. Now click the "Next" button to begin the installation. An installation progress bar appears so that you can follow the installation. Note that DualChip evaluation software requires that Java Runtime Environment (JRE) 1.3 has been installed on the computer. During installation, DualChip evaluation software determines whether the Java 1.3 Runtime Environment (JRE 1.3) needs to be installed. If the DualChip evaluation software cannot locate the proper runtime version, you will be prompted to install Java before continuing the installation. This installation is carried out automatically by the installation program. The DualChip evaluation software installation will automatically proceed after Java is installed. When the installation is complete, a message box appears. Click on Finish to complete the installation and close the install program. 4. Compatibility of DualChip evaluation software with quantification software 5

7 DualChip evaluation software can load data in tabulation delimited text format from the following quantification (feature extraction) software: 1.) GenePix Pro from Axon Instruments, Inc. 2.) QuantArray Analysis Software from PerkinElmer, Inc. 3.) ImaGene from Biodiscovery, Inc. 4.) ArrayVision from Imaging Research, Inc. 5.) Array-Pro from Media Cybernetics, Inc. The GeneID list corresponding to the used DualChip type (provided by Eppendorf in the DualChip kit) must be loaded by the user when quantifying. The user cannot make any changes to the GeneID. The software checks whether the used GeneID in the quantification file and chosen GeneID from the DualChip evaluation software are compatible. Note: Some of this software enables the quantification of several images at once, and saves the quantification data in a single file. This procedure can be used if images coming from experiment and reference are not quantified at the same time. The same order for reference and experiment must be used. You must generate one quantification file per dye and sample used. This procedure can be used if quantified files do not mix experiment and reference. A generic tabulation delimited text file loader is launched if another quantification software is used (section 6.2.2). 6

8 5. Data Analysis Workflow 5.1. Introduction DNA microarrays are powerful tools for the analysis of differences in the gene expression levels of a multitude of genes in parallel. In microarray experiments, hybridized microarrays are imaged using a microarray scanner to produce fluorescence intensity measurements that collectively cover the array. These intensities correspond to the expression level of each gene of the biological sample. Common sources of variation within microarray systems are varying RNA quality, varying cdna synthesis efficiency, dye-bias when using two-color fluorescence labeling systems, inconsistent hybridization conditions due to poor temperature control, inadequate mixing or distribution of the hybridization solution. The data analysis workflow implemented in the DualChip evaluation software is designed to minimize the effect of these sources of variation on the gene expression profile by using a special two-step normalization (section 5.6). As the primary objective of most microarray applications is to detect differences in gene expression levels, variations due to the technology should be kept to a minimum. The data workflow can be divided into several steps as highlighted in Fig. 1. Fig. 1: Workflow of a typical microarray analysis 7

9 The Eppendorf DualChip microarrays are most suitable for this data analysis procedure (Fig. 2). Fig. 2: Design of the DualChip human general. Internal Standards (yellow and orange; contain 6 differently concentrated plant-sequences) and house keeping genes (blue spots) are distributed over the array. All probes on the DualChip microarrays are spotted in triplicate. For differential gene expression analysis, the measured data of each gene of the experiment sample and of the reference sample are used and compared in terms of ratios (experiment/ reference). However, analysis of microarray data is a complex procedure involving several steps that may alter the quantitative aspect of the analysis. Each step of the entire analysis is accurately described below, starting from the raw data and continuing through the calculation of the gene expression ratios. The first step of the data analysis workflow consists in the extraction of the relevant signal intensities of each gene for experiment and reference data. This step consists of local background subtraction, subtraction of mean of signal intensities of negative controls, analysis of replicates and removal of replicates considered to be outliers. 8

10 Finally, the focus of this section is to explain how multiple scans of the arrays can improve the quality of the results Microarray control probes The extensive set of control probes printed on the DualChip microarray is an important feature for ensuring the generation of high quality results. The control set allows verification of cdna synthesis efficiency and hybridization as well as signal linearity. The control set is also intended for normalization of generated ratios. For more information on how to use the controls in the experimental procedure, please refer to the DualChip manual. This set of controls includes positive hybridization controls, negative hybridization controls, positive detection concentration curve, negative detection controls, an Internal Standard (mix of spiking RNAs), and a set of house keeping genes. These controls are intended to control and correct the possible variables occurring during the experiment. They make it possible to validate the data acquired at the end of the analysis. The following table reviews each of the controls and their purpose. Control name Type of molecule/ step Usage Internal Standard (mix of spiking RNAs) In-vitro transcribed polyadenylated RNA spiked into RT RT (Reverse transcription) control, hybridization control, and normalization Positive hybridization controls Biotinylated DNA spiked into the hybridization mix Hybridization control, grid alignment (indirect labeling) Negative hybridization controls Plant DNA on DualChip Detection of non specific hybridization (subtracted from signal) Negative detection controls Buffer on DualChip Check of quality of detection Positive detection concentration curve Biotinylated capture probe on DualChip House keeping genes (HKG) DNA Normalization Detection control and for grid alignment (indirect labeling) The controls enable the validation of the different steps of the analysis as summarized in the following table. Positive detection Internal Standards Positive hybridization Interpretation control * controls ** All steps are successful The cdna synthesis or hybridization step has failed or omission of key reagent The cdna synthesis does not occur or omission of key reagent The detection step has failed or omission of key reagent. * only valid for indirect labeling protocol ** for direct labeling only when using the optional hybridization control (+) = positive signal (-) = negative signal 9

11 5.3. Raw data transformation Hybridized microarrays are imaged by a microarray scanner. The fluorescence intensities of the whole microarray, including the hybridized probes, are typically stored as 16-bit images (tif-file) and described as "raw data". 1. A spot mask (grid) on the image allows the precise definition of the spot-location and the calculation of the fluorescence intensities of each spot. During this step, the classification of pixels as foreground (signal) or background will be carried out using dedicated software. 2. Afterwards, quantification can be performed to extract raw data intensities from the signal areas and their related background (Fig. 3). The image information is transformed into intensity values by this quantification-process. The quantification software generates a file containing the raw data for each spot on the array. The file format is specific to the software used. These files are imported into the DualChip evaluation software (sections 6.2.1, 6.2.2) for gene expression profiling. Fig. 3: (a) Original image of one spot of a DualChip microarray. (b) Segmentation of the spot. The green area displays the background signal and the red area represents the foreground signal. (c) Overlay of the original (a) and segmented image (b). (d) Spot is delimited by a red contour (spot mask/ grid). 3. It is essential in microarray image analysis to adjust for background signals. There are two different kinds of background that have to be subtracted from the signal. These subtractions are performed automatically by the DualChip evaluation software. To calculate the local background intensity of each spot (fluorescence emitted from other chemicals on the glass), a small region surrounding the spot mask has to be considered. The mean of background pixel values is subtracted from the total signal intensity of each spot. Additionally non specific binding signals, e.g. the mean signal of negative hybridization control spots (NHCS), is subtracted from the signal of every spot. Negative hybridization controls spots taken into account for the computation of this nonspecific fluorescence are selected using the following automatic procedure: 1. For all NHCS, the difference between signal and background (S-B) is computed 2. All NHCS are used as start SET 3. Steps 3 to 7 are used until no selection changes appear 4. For all NHCS in SET, steps 5 and 6 are repeated 10

12 5. Mean (M) and standard deviation (S) of all NHCS (S-B) in SET are computed without the current NHCS 6. Current NHCS from SET if (S-B < M-5S) or (S-B > M+5S) is removed. The purpose is to remove NHCS that are not similar to the other currently selected NHCS 7. Go to step 3 This procedure recursively removes the negative hybridization controls by comparing each control to the other selected controls in a T-Test-like procedure Data selection and replicates data merging In some microarray experiments, some spots may show scratches, dust particles or other irregularities may occur (Fig. 4). These unusual spots must be marked in order to make decisions about them at the time of data analysis. This process, carried out manually or automatically by the software, is called spot flagging. Fig. 4: Replicates are used to correct problems that may occur on the array, such as a hybridization problem in the lower right corner. Potential errors such as bubbles or dust may alter the signal intensity of spots (Fig. 4). The presence of spot replicates provides valuable quality information and allows the performance of statistical analysis and the correction of the gene data within the single experiment. Therefore, all genes of the DualChip microarrays are spotted in triplicate. By using three replicates you are less vulnerable to irregularities on the array. Having triplicates, spots presenting an abnormal variation of signal intensity are considered as outliers. They are removed. In addition, the coefficient of variation (CV = standard deviation/ mean) of the replicate signal intensities must be determined after the image has been quantified. 11

13 The following rules are applied to remove the outliers: - If the CV is higher than 0.7 and one replicate signal intensity is above 1,000, the replicate farthest from the mean is removed. - If the CV is higher than 0.3 and one replicate signal intensity is above 10,000, the replicate farthest from the mean is removed. These rules make it possible to discard a potential outlier. Due to the greater variability of low signals, the acceptance range for those genes is chosen to be broader. Following this removing process, the mean of remaining replicates (at least two replicates) will be computed Data validation Finally, a procedure for validating the intensity of spots is performed. Spots showing fluorescence intensity that is not in the linear range of measurement, e.g. either saturated or very low signal intensities, are sorted out. Saturation has a compressing effect on the ratio, and extremely low signal intensities yield higher variability in results. All genes not declared as saturated or low can be considered quantifiable for reliable ratio data generation. The validation procedure is as follows (section 6.9): For validation of single spot intensity: 1. If the intensity of the spot (S) is above a fixed threshold (by default 50,000), the signal intensity of the spot is saturated. 2. If the intensity of the spot (S) is less than its local background (B) multiplied by a fixed value (by default 2), the signal intensity of the spot is designated as too low (not detected). This can be expressed with the following formula: S < B x 2 3. The following rule compares the intensity of each spot to the intensity of the negative hybridization control spots after having subtracted their corresponding local background. If the intensity of the spot (S - B) is less than the mean intensity of negative hybridization controls (NHCS - B) (section 5.3) plus the standard deviation of the signal intensity of the negative hybridization controls (minus background) multiplied by a fixed value (by default 2), the signal intensity of the spot is designated as too low (not detected). This can be expressed with the following formula: (S-B) < [(NHCS - B) Mean + ((NHCS - B) Std dev) x 2)] In other cases, the spot is designated as quantifiable. The validation of mean gene intensity (mean of replicates) is performed in the same manner as the validation of single replicate intensity. A gene is designated as quantifiable when all the selected replicates are quantifiable. If not, the gene is designated as too low or saturated according to the replicates validation. 12

14 In the next step, gene expression ratios consisting of the signal intensities of experiment and reference are calculated for each gene (Fig. 5). If the signal intensity for both the experiment and the reference are quantifiable, a ratio is termed quantitative. If only one intensity is quantifiable (experiment or reference), the ratio is said to be qualitative. If neither intensity is quantifiable, the ratio is designated as being "not in linear range". Mean Signal Intensity of Replicates Resulting Ratios are: quantitative (quant.) qualitative (qual.) not in linear range (n.i.l.r.) too low quantifiable saturated Fig. 5: Graphical display of data defined as quantitative, qualitative, or not in linear range 5.6. Normalization The Eppendorf DualChip microarrays have been designed to efficiently integrate a two-stepnormalization (De Longueville et al. Biochemical Pharmacology 2002; 6: ). The kit includes an Internal Standard mix (a dilution series of in vitro-transcribed, polyadenylated RNA of six Lycopersicon esculentum genes 1 ), as well as the corresponding capture probes already printed on the DualChip microarray. Additionally, capture probes corresponding to numerous house keeping genes have been spotted throughout the array to allow a reliable two-stepnormalization. 1 The six Lycopersicon esculentum genes were selected because they are involved in plant-specific pathways and do not have any known mammalian homologues in the NIH sequence database. 13

15 The normalization process is summarized in Fig 6. Fig. 6: Uncorrected ratios are first normalized by a factor based on an Internal Standards and then by a mean house keeping factor Normalization based on Internal standard DualChip microarrays are divided into six zones, each containing two different concentrations of Internal Standard (IS) capture probes. These IS capture probes correspond to differentially concentrated RNA spiking standards (low and high) of the Internal Standard mix to ensure the acceptable expression of at least one internal standard per zone. This special design allows the computing of a local normalization factor for each zone using the quantifiable Internal Standard signal intensities. This local normalization factor is calculated from the intensity ratios of the Internal Standards of the corresponding zone in the reference and experiment samples. After calculating the local normalization factor, the ratios for each gene between experiment and reference samples in that local zone are corrected by the computed local normalization factor. Note: The kit includes an Internal Standard (several artificial spiking RNAs) for RT control, hybridization control, and normalization. First time users should familiarize themselves with the use of these controls before starting. 14

16 Normalization based on house keeping genes Internal Standards (exogenous RNA spikes) do not reflect the purity, quality, and amount of the analyzed RNA. Therefore, a second normalization step, based on expression levels of specific house keeping genes, is necessary. A house keeping normalization factor is calculated using the mean of quantifiable house keeping gene ratios Example A simple example of the necessity of normalization is presented in Fig. 7. Different amounts of starting material have intentionally been used for target preparation and a subsequent DualChip experiment to illustrate the normalization effect. The reference sample contains two times more RNA than the experiment sample. The scatter plot shows that signal intensities of the reference sample are higher than signal intensities of the experiment samples. The second step of the two-step-normalization procedure eliminates the different amount of starting RNA. Raw experiment data Normalized experiment data Raw reference data Normalized reference data Fig. 7: Raw data (a) based on hybridization of 10 µg (reference) and 5 µg (experiment) of rat liver RNA on DualChip rat hepato. Raw data (a) is clearly apart from the diagonal. The reference intensities are higher than the experiment intensities (ratio < 0) due to differences in the amount of starting material. (b) The normalization data for reference and experiment are comparable after two steps. (HKG = house keeping gene) 5.7. Statistical significance of gene expression ratios The DualChip evaluation software uses a statistical test for the evaluation of the statistical significance of gene expression levels. The aim of this step is to determine whether gene expression levels differ significantly between experiment and reference samples or whether the differences are due to random variation. The used model assumes the intensities to be Gaussian, independently distributed with the constant 15

17 coefficient of variation. Based on these assumptions, formulas were developed for the coefficient of variation and the confidence intervals. The coefficient of variation is estimated based on the subset of the ratios of house keeping genes selected for normalization that are assumed to be stable between experiment and reference samples after the house keeping gene normalization. This process is based on the assumption that the variation of house keeping gene ratios is representative for the variation of non-regulated genes. Therefore, the selected house keeping genes should have ratios distributed around the value 1. The significance of the ratios is established using the confidence interval computed from the statistical model: ratios outside the 95% confidence interval are statistically significant. Ratios outside the 99% confidence interval are statistically highly significant. Furthermore, the ratios of analyzed genes (e.g. coming from experiment divided by reference) can be grouped into (semi-)quantitative and qualitative. Ratios involving two quantifiable signal intensities are considered as quantitative. Ratios involving one low and/ or one saturated intensity are considered as qualitative. Ratios resulting from two low or two saturated intensities are designated as "not in linear range". Note: variability computed on the selected house keeping genes are limited by two values. This means if the computed variability is below the lower limit, the variability used for the calculation of the confidence interval is set to this lower limit. The same process applies for the upper limit if the computed variability is above the upper limit. This process is applied in order to minimize the possibility of obtaining false positives (e.g. non regulated genes are called regulated) and false negatives (regulated genes are called non-regulated). The result of these thresholds is that no ratio below is determined to be significant and all ratios above are defined as significant (1/ and 1/ for down regulated genes) Multiple scanning data As mentioned before, quantitative data analysis can only be performed if the signal intensities of the experiment and the reference are in the quantifiable range corresponding to the linear part of the scanning curve. Only genes that are not saturated or not very low (lower than 2 times the local background) can be used for quantitative data analysis. To increase the dynamic range of the system, and thus also the number of quantifiable genes, we recommend scanning DualChip microarrays with three different PMT settings. To do this, an intermediate PMT setting is first chosen to analyze medium expressed genes. Highly expressed genes can be accurately quantified at a low PMT setting and low expressed genes at a high PMT setting. An example of the same array scanned at different PMT settings is shown in Fig

18 (a) (b) (c) Fig. 8: Example of one array scanned at different PMT settings: (a) array scanned at high PMT setting; (b) array scanned at medium PMT setting; (c) array scanned at low PMT setting. The entire data analysis procedure shown in Fig. 1 is used independently for each scanner setting. Both normalization factors, the value of Internal Standards and the value of house keeping genes, are calculated for each scanner setting using quantifiable data from all the scans. Finally, each gene will produce as many ratios as PMT scans used. We recommend using only the ratios from quantifiable intensities whenever possible in order to obtain quantitative data. Based on several examples, the use of the different ratios to obtain a reliable ratio for each gene. Basically, the best way is to use only quantifiable values. If ratios generated from quantifiable intensities are available, the means of these ratios are used. If no quantitative ratio is available, qualitative ratios are used. To summarize, quantitative results overrule qualitative results and qualitative results overrule "not in linear range" results. Tab. 1: Summary of 3 PMT setting ratios for several genes using typical ratios Gene 1: 2 ratios are quantitative; mean of the 2 ratios is used for overall ratio Gene 2: 1 ratio is quantitative, 2 are not in linear range; the quantitative ratio is used for overall ratio Gene 3: 1 ratio is quantitative, 1 is qualitative and 1 is not in linear range; the quantitative ratio is used for overall ratio 17

19 Gene 4: 2 ratios are qualitative, 1 is not in linear range; mean of qualitative ratio is used for overall ratio Gene 5: 1 ratio is qualitative, 2 are not in linear range; the qualitative ratio is used for overall ratio Gene 6: 3 not in linear range ratios: mean of the 3 ratios is used for overall ratio 6. Work procedure 6.1. Step 1: Mode 1: DualChip evaluation software launching To start DualChip evaluation software, either: double click on the DualChip evaluation software icon on the Windows desktop or use the start menu: Start > Programs > DualChip evaluation software > DualChip evaluation software 6.2. Step 2: Mode 1: Quantification data for reference and experiment loading How to load data from supported quantification software To load the quantification data from supported quantification software for one experiment and reference comparison carry out the following procedure: 1. Click "Add files(s)" under the "Experiment Files" text box 2. Select the file(s) corresponding to the quantification data of the experiment (use Ctrl or Shift for multiple selections) 18

20 3. Repeat 1. and 2. for the reference file(s) Note: order of files in the experiment/ reference file display must match between experiment and reference files. Ratios are computed separately from scanner setting to scanner setting following the data analysis workflow explained in the previous section. It is preferable to load files in the following order: low gain, medium gain, and high gain. 19

21 How to load data from unsupported quantification software: generic text file loader If you use an unsupported software for quantifying the experiment and reference, please follow the steps described in the sections 6.2.1, and 6.3. Click the "Analyze" button to launch the analysis. The following window appears. To load the data in the DualChip evaluation software correctly, please enter the five requested parameters (parameters visible when opening the.txt-file in MS-Excel): Delimiter symbol: First row of data: Name column/ GeneID column: Signal column: Background column: symbol separating two values in the data files row in the data files containing the first spot data. Typically, this row is below the file header and behind the name of the columns column containing GeneIDs or any identifier for each spot. These GeneIDs must be loaded at quantification time column containing the signal quantified data column containing the background quantified data Note: all files to be loaded must have the same parsing parameters. The files should end with the latest data row and nothing should follow this row: no "END DATA" tag or empty lines for example. The GeneID contained in the files should not be surrounded by single or double quotes Load GeneID file GeneID files contain information about the location of each gene on the DualChip, the area of the genes for the Internal Standard normalization, and the gene accession number for linking the genes to databases. These files are obligatory for analysis. The GeneID file for the DualChip evaluation software corresponding to the GeneID file used in quantification software must be loaded. Please click specific GeneID file.. A window will pop up. Please select the 20

22 GeneID files are provided on the DualChip evaluation software CD. If the DualChip evaluation software GeneID file required for your analysis is not available, you can obtain updates concerning the DualChip microarrays by visiting the following page of our website When the analysis is launched (see next step), the DualChip evaluation software compares the GeneID used for quantification (for every quantification file loaded) with the GeneID file loaded in the DualChip evaluation software. If there is no difference, the analysis will continue. If there is a difference in at least one GeneID, the following window will appear: This difference can occur when either the GeneIDs are missing in the quantification file, when the orientation of the GeneIDs used for quantification and for the DualChip analysis software are not the same or when the GeneID file loaded in the DualChip evaluation software does not correspond to the type of DualChip analyzed. If you forgot to include the GeneID file at quantification time, and if you are sure of the orientation and DualChip type you are using, you can proceed by clicking "OK". The GeneID file stored in the DualChip evaluation software will be used for the analysis. In any case, the DualChip evaluation software compares the number of spots in the quantification files and the number of spot in the GeneID file. If they are not equal, a warning will appear. An example of such a window is displayed hereafter: 21

23 The analysis is then cancelled Step 3: Mode 1: data analysis launching The data analysis workflow implemented in the DualChip evaluation software is based on the DualChip array design and the DualChip protocol. The data analysis workflow is divided into several steps described in the previous section. Click to start the analysis. In the first step of the data analysis workflow, the relevant signal intensities of each gene for experiment and reference data are extracted. This step consists of local background subtraction, subtraction of negative controls mean, analysis of replicates, and removal of replicates considered as outliers (described in 5.4.) Step 4: Mode 1: Internal Standards normalization One of the important features of all DualChip microarrays that ensures the generation of high quality results is the extensive set of control probes printed on the microarray. The control set makes it possible to check the efficiency of cdna synthesis, hybridization, and signal linearity. Furthermore, the control set is intended for normalization of generated ratios. The set of control probes is completed by the Internal Standards (mix of spiking RNAs), which are added to the cdna synthesis reaction and used as positive control for cdna synthesis and for normalization of the results (section 5.6). The special DualChip design allows computing of a local normalization factor for each area using the quantifiable Internal Standard intensities of the area. The ratios for each gene between experiment and reference samples are first corrected by these local normalization factors (ratios), which are calculated from the intensity ratios of the Internal Standards in the experiment and reference samples. 22

24 Internal Standards table After starting the analysis, a window is displayed to draw the users attention to the importance of the Internal Standards normalization (IS normalization). A table summarizing the internal standards ratios between experiment and reference intensities is displayed. This table is comprised of the following columns: - Gene ID: GeneID of the Internal Standards - Zone: Normalization zone of the Internal Standard (section 5.6) - Overall ratio: Mean of acceptable PMT ratio - Ratio 1,, Ratio n: Each separate PMT ratio Internal Standards are automatically selected based on the acceptability (section 5.5) of PMT ratio. All genes comprising at least one acceptable PMT ratio are automatically selected. Internal Standards without any acceptable PMT ratio have an overall ratio set to "NaN". This particular value means "Not a Number". These Internal Standards should not be selected in any case Color code Selected Internal Standards are highlighted with a gray background and white font. For the overall ratio column: : Unselected Internal Standards that are part of the automatic selection (at least one acceptable PMT ratio) are highlighted with a green background and black font : Unselected Internal Standards that are not part of the automatic selection are highlighted with a white background and black font and "NaN" value. 23

25 For PMT ratio columns: : Unselected Internal Standards with an acceptable ratio are highlighted with a green background and black font : Unselected Internal Standards with an unacceptable ratio are highlighted with a white background and black font Selection Note: Internal Standards are pre-selected by DualChip evaluation software. Normalization is performed according to a data selection algorithm. Sometimes, it might be necessary to change the mathematical selection proposed by DualChip evaluation software. In this case, select alternative rows in the table by clicking on them (use Ctrl or Shift for multiple selections). Original selections made by DualChip evaluation software can be restored by using the Restore default selection button. Right clicking on a cell will display the following menu for restoring the selection or unselecting all the Internal Standards. Do not select Internal Standards having a NaN overall ratio Further interaction The value of the cell can be displayed in the bottom window when the specific cell of the GeneID, zone or overall ratio is marked by the mouse. By moving the mouse over a cell of the PMT ratio columns, the signal and background intensities for each spot of the corresponding gene and PMT settings are displayed. 24

26 To sort the table according to special column in ascending order, please click on the header of that column. Clicking on the header of a column while pressing the shift key sorts the table according to this column in descending order. For normalization using these chosen Internal Standards, please click. 25

27 6.5. Step 5: Mode 1: House keeping genes normalization Internal Standards do not correct differences in the quality and amount of starting material. This can be corrected in the second step: a normalization factor is computed using the mean of acceptable house keeping gene ratios. As it becomes clearer that there is more variation in transcription levels of house keeping genes than initially thought, selection of the house keeping genes for the normalization becomes a delicate decision. Only those house keeping genes whose expression levels were not changed between samples may therefore be used (section 5.6) House keeping gene table After the normalization with Internal Standards, a window is displayed to draw the attention of the user to the importance of the house keeping genes normalization. A table summarizing the house keeping gene ratios between experiment and reference intensities is displayed (page 25). An automatic selection of the house keeping genes is proposed by the software. This automatic selection assumes that the quality and amount of starting material are similar for experiment and reference samples. The automatic selection is performed with the following algorithm. The purpose of this algorithm is to compare each house keeping gene to the other house keeping genes using the statistical model described in section 5.7. The algorithm is divided in the following steps: 1. All house keeping genes with acceptable ratios are selected to form an initial SET 2. Steps 3 to 6 are repeated until no house keeping gene is removed from SET 3. For each house keeping gene in SET, steps 4 to 5 are repeated 4. Based on SET where the current house keeping gene has been removed, the confidence interval is computed based on the statistical model described in section If the house keeping gene ratio is outside the confidence interval, the current house keeping gene is removed from SET. It is otherwise kept. 26

28 6. Go to step 2 At the end of the process, housekeeping gene ratios with at least one acceptable ratio and for which all acceptable ratios are inside the confidence intervals are selected Color code Selected house keeping genes are highlighted with a gray background and white font. For the overall ratio column: : Unselected house keeping genes that are part of the automatic selection (at least one acceptable PMT ratio but with all acceptable ratios inside the confidence interval) are highlighted with a green background and black font. : Unselected house keeping gene that are not part of the automatic selection are highlighted with a white background and black font. For PMT ratio columns: : Unselected house keeping genes with acceptable ratios inside the confidence interval are highlighted with a green background and black font. : Unselected house keeping genes with acceptable ratios outside the confidence interval are highlighted with a green/ blue background and black font. : Unselected Internal Standards with unacceptable ratios are highlighted with a white background and black font. 27

29 Selection Note: House keeping genes are pre-selected by DualChip evaluation software. Normalization is performed according to a data selection algorithm. In some cases it might be necessary to change the mathematical selection proposed by DualChip evaluation software. In this case select alternative rows in the table by clicking on them (use Ctrl or Shift for multiple selections). Original selections made by DualChip evaluation software can be restored by using the Restore default selection button. Right clicking on a cell will display the following menu for restoring the selection or unselecting all house keeping genes. Do not select genes having a NaN overall ratio Further interaction Clicking on the header of a column sorts the table according to this column in ascending order. Clicking on the header of a column while pressing the shift key sorts the table according to this column in descending order. By moving the mouse over a cell of the GeneID, zone or overall ratio column, the value of the cell is displayed in the bottom window. By moving the mouse over a cell of the PMT ratio columns, the signal and background intensities for replicate spots of the corresponding gene and PMT setting are displayed. See following picture. 28

30 For the second step of normalization, using chosen house keeping genes, please click. 29

31 6.6. Step 6: Mode 1: Gene expression profile tables: report of the analysis Ratio tables At the end of the analysis, the results are summarized in a table. 30

32 This table contains the following information about every gene for each PMT setting used in the analysis: GeneID: gene names corresponding to each spots Zone: array zone where that particular gene is located Unnorm. Ratio: ratio of corrected signals between experiment and reference before normalization. Norm. Ratio: ratio of corrected signals between experiment and reference after normalization. Status: qualitative, quantitative, or not in linear range. Genes are flagged according to the relevance of the ratio. Quantitative ratio means the experiment and reference signals are acceptable (not saturated and detected above the background). Qualitative ratio means either the experiment or the reference signal is acceptable. Not in linear range ratio means neither the experiment nor the reference signals are acceptable. Significance: statistical analysis of the significance of the ratio compared the ratios of selected house keeping genes (Unchanged, Significant, Highly Significant). Ratios are compared to confidence intervals computed mathematically. Gene ratios inside the 0.95 confidence interval can not be considered statistically different from 1. They are flagged as unchanged. Ratios outside the 0.95 confidence interval are flagged as significant. Ratios outside the 0.99 confidence interval are flagged as highly significant. Genbank: Genbank accession number Swissprot: Swissprot accession number Comments: DualChip evaluation software issues comments. The comments for both the experiment and reference are of 3 types: comment on the acceptability of the intensity: OK if the signal is acceptable, "Low signal" means that the signal is too low and "Saturated signal" means that the signal is above the setting for saturation (sections 5.5, 6.9). Positive hybridization controls can have saturated signals under normal conditions. For this reason, their comment is OK, even if their signals are saturated. comment on the coefficient of variation (CV) of the replicates used for the gene. If the CV is higher than the corresponding parameter (by default 0.3, section 6.9), the comment "High CV" is displayed. Comment on the signal to noise ratio is below the corresponding parameter (by default 1.5, section 6.9), the comment "Low SNR" is displayed. 31

33 An overall profile table is also computed by combining the different scanner settings tables (section 5.8). For this table, the comments column is empty. To navigate through these tables, click on the thumbnails shown in the following picture Color code GeneID column: : All the controls are highlighted with a green background and black font. See section 6.9 for the list of all controls. : Genes of interest are displayed with a white background and black font. Significance column: : All highly significant (section 5.7) genes are highlighted with a green background and black font. : All significant (section 5.7) genes are highlighted with a dark green background and white font. : All other genes are displayed with a white background and black font. Comments column: : Genes with acceptable ratios are displayed with a white background and black font. 32

34 : Genes for which the reference or experiment are saturated are highlighted with a pink background and black font. : Positive hybridization controls with saturated intensities are highlighted with a pink background and black font. : Genes for which the experiment or reference have a low signal mean or low signal to noise ratio are highlighted with a pale blue background and black font. (The case saturated signal and low signal mean is marked in red.) Normalization data and negative hybridization controls mean Click the "Norm. data" thumbnail to display normalization data and negative hybridization controls information. The table is divided into two parts. The first two columns are dedicated to the normalization ratio used (sections 5.6, 6.4, 6.5). The second part displays the negative hybridization mean (CM) and standards deviation (SD) for each scanner setting (section 5.3). The normalization table gives the Internal Standards normalization ratio for each of the six zones of the DualChip microarray. The last row gives the computed normalization factor based on the selection of house keeping genes. 33

35 Each row of the negative hybridization table is composed of three cells: - Channel: the quantification file (exp. for experiment and ref. for reference) - CM: Mean of the negative hybridization controls. - SD: Standard deviation of the negative hybridization controls Further interaction Clicking on the header of a column sorts the table according to this column in ascending order. Clicking on the header of a column while pressing the shift key sorts the table according to this column in descending order. Click text files. to choose the result file name and save the tables content in tabulation delimited 34

36 Note: If you want to import the report-txt-file into MS-Excel for further data analysis, please be aware of using English-(US)-settings. Otherwise the comma/ dot tabulation can be mixed up Step 7: Mode 1: scatter plots and histograms Click on the "Plots" thumbnail to access scatter and bar plots of the data. Scatter plots for unnormalized and normalized data are available for each gain. Click on the corresponding gain button to display the plot. Move the cursor over a gene to display a window containing gene name and signal intensities for this gain. Note: a scatter plot is not available for the overall data because the data cannot be directly expressed as the ratio of experiment signal over reference signal. 35

37 In the plot there are several markings: - Black circles correspond to unchanged ratio. - Dark green circles correspond to significant ratio. - Light green circles correspond to highly significant ratio. The "select" and "Lasso" tools permit the selection of ratios either by rectangle selection or by free selection respectively. To zoom in on a rectangular section of the scatter plot, select the zoom tool and then the area. A right mouse click can be used to cancel the last zoom. Bar plots of the ratios are also available for each gain. 36

38 The "select" tool permits the selection of ratios by rectangle selection. To zoom in on a rectangular section of the scatter plot, select the zoom tool and then the area. Right clicking cancels the last zoom step. 37

39 6.8. Load saved report The DualChip evaluation software enables the loading of the report from an analyzed gene expression experiment. Click the button to load the saved text report file. The software locates the quantification and the GeneID files used for the analysis. If one of these files has been moved and saved in another directory, the software will be unable to load the report file. The locations of the quantification and GeneID files are described in the header of the result file. One of the following windows will appear: To repeat the normalization with the stored Internal Standards and house keeping genes, click "yes". The results of the analysis will be displayed (section 6.6) Settings - Parameters The default setting parameters are verified for the DualChip-concept. However, they are changeable. The settings are accessible with the menu file > settings. The settings are separated under three thumbnails. The first page (Measurements) is dedicated to the choice of the signal and background type of the intensity characteristic: median or mean. The default values are the median for the signal and the mean for the background. 38

40 The second page (Controls) displays the seven types of control spots. These parameters should not be changed. They are linked to the GeneID files and permit the recovery of the different control spots. 39

41 The third page (Quality Control) enables the setting of the parameters for the removal of the outliers in the triplicates used in the data analysis software. This function differentiates the too low, acceptable and saturated spots and comments the genes based on the coefficient of variation and signal to noise ratio. The default values are displayed in the following picture Step 8: Mode 2: Analysis of experiment replicates The second mode of the DualChip evaluation software is devoted to the analysis of replicates of gene expression analyses with several experiments and references. The purpose of this mode is to obtain relevant mean ratios for each gene by combining ratios obtained with several replicates of the same gene profiling experiment. This mode also enables the evaluation of the variability between replicates of the same experiment. To load data, please click "Add File(s)". 40

42 and select tabulation delimited text files obtained after the analysis with the first mode of the DualChip evaluation software for each experiment replicate (use Ctrl or Shift for multiple selections). 41

43 Click to perform the analysis Automatic selection. A table with an automatic selection of the ratio to be taken into account for each gene is computed based on the status and significance of the ratio. This automatic selection is based on the following algorithm: Determine the majority class based on the status: quantitative, qualitative or not in linear range. In case of impasse, quantitative stands above qualitative and qualitative stands above "not in linear range". Among the majority class ratios determined previously, determine the majority class based on the significance: highly significant, significant, or unchanged. In case of impasse, highly significant stands above significant and significant stands above unchanged. The ratio selects all the ratios of the majority category for each gene. An example of the results table is displayed in the following picture: 42

44 Color code : Quantitative unchanged categories/ ratios are highlighted with a pale blue background and black font. : Undetected categories/ ratios are highlighted with a white background and red font. : Quantitative significant categories/ ratios are highlighted with a green background and black font. : Qualitative significant categories/ ratios are highlighted with a dark green background and white font. 43

45 : Unselected ratios are highlighted with a gray background and dark gray font Selection change Automatic selections can be changed by selection of a cell from the table. A right mouse click will display the following menu. Choose one of the options to change the selection. The overall ratio, coefficient of variation and category are updated according to the new selection. The category is the majority category as explained in the previous section Further interaction Clicking on the header of a column sorts the table according to this column in ascending order. A click on the header of a column while pressing the shift key sorts the table according to this column in descending order. Moving the mouse over a cell of the GeneID, overall ratio and CV displays the value in the bottom window. Moving the mouse over a cell of the experiment ratio column displays the intensities of the signal and background for all the spots of all the PMT settings used for the experiment and the reference. All the ratios of the different PMT settings are also displayed. 44

46 To save the table in tabulation delimited text file, click Color code summary Internal standard table Selected internal standards Overall ratio column Overall ratio column PMT ratio columns PMT ratio columns At least one acceptable PMT ratio No acceptable PMT ratio Acceptable ratio Unacceptable ratio House keeping genes table Selected house keeping gene Overall ratio column Overall ratio column PMT ratio columns PMT ratio columns PMT ratio columns At least one acceptable PMT ratio No acceptable PMT ratio Acceptable ratio inside confidence interval Acceptable ratio outside confidence interval Unacceptable ratio GeneID column Report ratio table Controls are highlighted GeneID column Gene of interest Significance column Significance column Significance column Highly significant gene Significant gene Other gene Comments column Acceptable ratio 45

47 Comments column Saturation in reference or experiment Comments column Comments column Positive hybridization controls with saturation Low signal in reference or experiment Category and ratio column Category and ratio column Category and ratio column Category and ratio column Ratio column Multiple experiment analysis Quantitative unchanged category/ ratio Not detected category/ ratio Quantitative significant category/ ratio Qualitative significant category/ ratio Unselected ratio 46

48 47