GEView (Gene Expression View) Tool

Download and Installation: http://www.weizmann.ac.il/complex/compphys/software/geview/

Note on Linux version: The result Excel file can be opened by OpenOffice with similar functionality like on Windows.

General Description:

This tool was developed by: Libi Hertzberg, Assif Yitzhaky (Domany group, Weizmann Institute of Science) and Metsada Pasmanik-Chor (Head of Bioinformatics Unit, Faculty of Life Science, TAU).

The following flow chart summarizes the functionalities of GEView.

Input:

Microarray format:

A text tab-delimited file with normalized and summarized expression values (log 2 scale). Each row (beginning from the 3rd row) represents a probe set for a specific gene, and each column (beginning from the 3rd column) represents a sample. 

·         The first column contains the probe set name.

·         The second column contains the gene symbol name. 

·         The first row contains the samples' individual names.

·         The second row contains the samples' labels. Labels are short and clear descriptions of the sample type, containing letters and numbers only, with no spaces. Underscores ( _ ) are used to separate between the types of labels. An example for a text representing the labels of one sample: CNT_Bt1_C. In this example Label 1 is CNT (control), Label 2 is Bt1 (batch1) and Label 3 is C (child). It means that this specific sample is from the control group, was measured in batch number 1, and is of a child. Label 1 represents the different conditions of the experiment, and will be used for the t-test or ANOVA and for grouping samples in the box plots. The number of labels (= [number of underscores (_) +1] is not limited, and the number of conditions inside each label is also not limited.

The data begin from the 3rd row and the 3rd column. 

Note on data scale: the expression values can be on log 2 scale (default) or raw data (in this case please make sure that all values are positive since log 2 transformation will be applied on your data by GEView.

In the following input example, the first level (the conditions for the statistical analysis) has either the value “CNT” or “DIS”. The second level (after the underscore  _ ) designates the batch (Bt1 or Bt2).

 

 

 

Sample1

Sample2

Sample3

Sample4

Sample5

Sample6

 

 

CNT_Bt1_M

CNT_Bt1_F

CNT_Bt2_F

DIS_Bt1_M

DIS_Bt2_F

DIS_Bt2_M

34689_at

TREX1

5.5

5.3

4.7

7.5

12

4.4

34697_at

LRP6

3.7

7.6

5.7

5.6

8.5

3.7

34726_at

CACNB3

9

5.9

4.6

7.6

8

6.7

 

         

          A small tab-delimited input file example can be found here: data_example_small.txt

Tip: In order to create a similar expression table, you can use the EXPANDER tool (http://acgt.cs.tau.ac.il/expander/) or the Expression console tool (for Affymetrix microarray data) http://www.affymetrix.com/estore/browse/level_seven_software_products_only.jsp?productId=131414#1_1 .

Next Generation Sequencing (NGS) / protein format:

Like the microarray format as described above but without the first column. In this case the only descriptive column is gene symbol name:

Example:

 

Sample1

Sample2

Sample3

Sample4

Sample5

Sample6

 

CNT_Bt1_M

CNT_Bt1_F

CNT_Bt2_F

DIS_Bt1_M

DIS_Bt2_F

DIS_Bt2_M

TREX1

5.5

5.3

4.7

7.5

12

4.4

LRP6

3.7

7.6

5.7

5.6

8.5

3.7

CACNB3

9

5.9

4.6

7.6

8

6.7

 

          GUI Layout – 3 stages:                

            

          Stage 1 – Load data

          Use the “Load” button in order to browse for the Data File. After choosing the Data File, its path will appear in the Data File text box.

Stage 2 – Preprocessing

Click “PCA” in order to plot the Principal Component Analysis of the samples based on (up to) the 1000 highest variance genes. Using the new figure toolbar, you can rotate (3D) the PCA figure, zoom in, save, etc.

If you right-click anywhere on the white space between the points (but not on the points themselves), a context menu will appear, enabling you to recolor the points according to any one of the various labels (as defined in the second row in the input file, see the input section above)

           

If you right-click on a point (sample), a context menu will appear, enabling you to remove (filter out) this point from the analysis.

          If you click “Remove Sample…” this point will be designated as “Removed”, and the sample name will appear in the stage 2 “filtered samples” list (see the figures below).

         

         

          After removing a sample, you can add it again (cancel the sample removal) by right-clicking again on the sample and choosing “Add Sample…”.

          After removing the unwanted samples, click “PCA” again in order to recalculate and redisplay the PCA of the desired samples.

Batch correction:

Use this button if you have various batches and you would like to perform batch effect correction. Note: the batch label must be the second level, after the first underscore (_) (second line in the input file), see the input section above.  Moreover, In order to run the batch correction, in each batch group, all the various conditions (first label level) must appear.

After the batch correction is completed, an additional PCA figure will open, reflecting the batch correction output.

The Batch correction algorithm utilized in GEView is ComBat. It is implemented in the sva (Surrogate Variable Analysis) package of Bioconductor.

https://bioconductor.org/packages/release/bioc/html/sva.html

References:

-         Johnson WE, Li C, Rabinovic A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics. 2007; 8(1):118–27. Epub 2006/04/25. doi: 10.1093/biostatistics/kxj037 PMID: 16632515.

 

-         Jeffrey T. Leek, W. Evan Johnson, Hilary S. Parker, Andrew E. Jaffe and John D. Storey (). sva: Surrogate Variable Analysis, R package version 3.6.0.

Batch correction Example:

Before: in the following figure, the samples are grouped according to batches

After: in the following figure, the samples are grouped according to the experiment conditions

Stage 3 – Run

Running the statistical analysis may take a while depending on the number of the processed probe sets. You can follow the running progress by watching at the status line at the bottom of the graphical interface.

This stage generates an Excel file which summarizes in each line a specific probe. The probe sets are sorted (starting from the most significant) by the FDR ANOVA Q-value (corrected p-value, see reference below). Each row contains a link for the figure containing the ANOVA box plots. Note: in case that there are more than two experiment conditions (more than two groups in the first level of the labels), then Tukey’s test for multiple comparison is performed; for each subgroup pair – the fold change (f.change) and the p-values (“p-val”) are given.

 Example for output table:

Probe set ID

Gene Symbol

GeneCards Link

UCSC Link

NCBI Link

ANOVA P-value

Q-value (corrected)

Box plots

E2A/HD f.change

p-val

E2A/TEL

f.change

p-val

HD/TEL

f.change

p-val

212148_at

PBX1

GeneCards

UCSC

NCBI

2.30E-15

3.80E-11

Figure

20.3

1.90E-09

21

1.90E-09

1.03

0.9

212151_at

PBX1

GeneCards

UCSC

NCBI

3.50E-13

2.90E-09

Figure

25.2

1.90E-09

20.6

1.90E-09

0.817

0.2

200953_s_at

CCND2

GeneCards

UCSC

NCBI

2.10E-07

0.00022

Figure

0.0787

1.70E-07

0.223

1.40E-05

2.83

0.00089

201005_at

CD9

GeneCards

UCSC

NCBI

1.30E-06

0.00055

Figure

0.874

0.92

10.7

6.00E-06

12.3

6.90E-06

205253_at

PBX1

GeneCards

UCSC

NCBI

3.70E-06

0.0011

Figure

10.8

2.60E-05

10.8

6.10E-06

1

1

204849_at

TCFL5

GeneCards

UCSC

NCBI

2.20E-05

0.0029

Figure

0.639

0.34

0.152

2.70E-05

0.238

0.00069

200951_s_at

CCND2

GeneCards

UCSC

NCBI

3.80E-05

0.004

Figure

0.372

3.10E-05

0.744

0.083

2

0.00049

221773_at

ELK3

GeneCards

UCSC

NCBI

0.0011

0.024

Figure

0.195

0.0016

0.279

0.0039

1.43

0.56

200952_s_at

CCND2

GeneCards

UCSC

NCBI

0.13

0.27

Figure

0.789

0.18

0.812

0.18

1.03

0.97

206127_at

ELK3

GeneCards

UCSC

NCBI

0.28

0.43

Figure

0.988

0.99

0.905

0.32

0.916

0.45

 

 

 

 

 

 

 

FDR reference: Benjamini, Y., and Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society 57, 289–300.

In case that a particular gene symbol has multiple probe sets, then all probe sets for the same gene are combined together in the same figure (see an example below). The figure displays 2 examples: A) 3 probe sets of PBX1. B) 3 probe sets of ELK3. The Y-axis represents the Log2 expression value and the X-axis represents the 3 experimental conditions (see E2A, HD and TEL X-labels). The number of samples in each experimental condition is written in parentheses. Each sample expression value is marked by a red dot and the mean (default) and standard deviation in each experimental condition are marked by a black line and a blue box, respectively.

 

          Run parameters

1.     You have the option to display in the figure either the mean (default) or the median value of each group of conditions. In the example below, the mean is shown as black line.

2.     The parameter “Display sample names in Fig.” determines which sample names will be displayed in the figure. The default value (min & max exp.) displays the names of only 2 samples in each box: the samples with the minimal and maximal expression. You can also choose to display all the samples (note that this may be overcrowded), or none of them. In the example above, no sample names are shown.

3.     The checkbox “only probe sets with gene symbols” (default is checked) determines whether all probe sets will be analyzed or only probes sets which have gene symbols (default).