Tutorial.2-channel

Carl Von Albert

Informations

• Publié par: Mushur
• Nombre de lectures: 6
• Langue: English

Extrait

VAMPIRE: Two-channel Microarray Tutorial

2005.02.17

This tutorial will take you through the typical steps needed to interpret a set of two-channel

microarray data. The data set used for this tutorial was obtained from an AfCS study on the effects of

LPS on a RAW264.7 macrophage cell line. We will use this set to explore the response of these cells

after 1 hour of treatment.

Sample Editor

The first step will be to load the data set into the VAMPIRE microarray analysis platform through the

sample editor. You will load the file “tutorial.2-channel.txt” as a

table

of measurements on the

AfCS16K

array. You may enter whatever description you wish. Note the file format of the tutorial file.

Lines that are blank or are preceded by the # character are ignored. The first interpreted line contains

the titles of each of the samples contained in the file. The first column contains the names of each

feature. Each successive column contains gene expression measurements for each sample.

Note: Data should

not

be log-transformed before loading into VAMPIRE. Normalized data from

RMA, CORGON, dChip, etc should be used with caution, as these tools have profound effects

on the variance structure, and can prevent the variance structure of the data from being

adequately modeled. For Affymetrix chips, we recommend MAS 5.0/GCOS scores. For Agilent

chips, we recommend the processed signal intensities.

Group Editor

Next, you will learn how to group related samples. For the paired analysis that we will be performing,

it is necessary to create not only groups for the two conditions that we wish to compare, but also

groups to pair samples that are measured on the same array.

Select the

AfCS16K

array type and create the following groups:

Group name

Contents

control

con1, con2

lps

lps1, lps2

array1

con1, lps1

array2

con2, lps2

The first two groups, “control” and “lps” are groups for the two treatment conditions that we wish to

compare. We would like to find out whether there is a systematic difference in gene expression

between control-treated macrophages and LPS-treated macrophages. The latter two groups, “array1”

and “array2” will be used to pair these samples together. Since measurements made on the same

microarray are likely to move together, it is important for the statistical method to recognize which

samples were measured off the same array.

Category Editor

Create a new category that contains all of the “pairing” groups. You will see the importance of this

step later on.

Select the

AfCS16K

array type and create the following category:

Category name

Contents

array

array1, array2

Variance Modeler

With all of the necessary groups and categories defined, you will need to create a paired variance

model to estimate the error structure of the data set that you’ve submitted. Create a model named

“conlps”. Select the

AfCS16K

array type, and create a “paired” model.

Select the following:

Sample group 1 – control

Sample group 2 – lps

Pairing category – array

Once you’ve clicked on “Build Model”, a job will be submitted to the database, which will be

processed in the order which it was received. If no other jobs are running, this step will typically take

about 20 minutes. In the meantime, we may go ahead and create the statistical test that we wish to run.

When the variance modeling job is completed, the model itself may be viewed by selecting it from the

navigation bar on the left. Parameter values and their MCMC simulation errors will be reported in the

results section of the page. In this section, there are a number of critical values that must be examined

to ensure the reliability of this analysis.

Check the following:

1. cutoff(quantile) should be a value greater than 0 and less than 0.8.

2. the A parameters should be similar between the two experimental conditions

3. the B parameters should be similar between the two experimental conditions

4. rhoA and rhoB will ideally be values greater than zero and less than 1

If any of these conditions are not true, the results of the significance test should be treated with caution,

because it is likely that the cutoff procedure was not able to identify a stable estimate of the variance

parameters. This can occur for a number of reasons:

1. the raw data was pre-processed with a complex normalization algorithm, introducing

normalization artifact

2. the raw data contains control probes that do not share the variance structure of the rest of the

data set (Agilent arrays)

Significance Tester

Finally, you will create a significance test based on the model that you created to identify which array

features are differentially-expressed between the two experimental conditions. Select the

AfCS16K

array type, and create a “paired” test.

Select the following:

Sample group 1 – control

Sample group 2 – lps

Pairing category – array

Variance model – conlps

After an additional 10 minutes, this job will complete and you will be able to download the results of

the VAMPIRE analysis. Simply select the significance threshold that you desire, and click download.

The results may be downloaded either as a tab-delimited text file or as a VAMPIREResults XML file.

GOby

GOby has two functions that may be useful for the interpretation of microarray data.

1. The “Annotate” function may be used to construct an annotated table of differentially-

expressed genes from multiple statistical tests.

2. Alternatively, you may wish to use GOby to identify functional groups that are

overrepresented in the list of differentially-expressed features. Click on “New” in the

navigation bar on the left. Then, simply select the test you wish to interpret, select a

significance threshold, and click “analyze”. The GOby analysis job will take approximately

30 minutes to complete. The results may subsequently be viewed as a series of HTML web

pages, or downloaded as GObyReport XML documents.