Biostar Beta. Not for public use.
Microarray data analysis
0
Entering edit mode
13 months ago
Adarsh Kuamr • 40

Dear friends, I am doing microarray data analysis by GeneSpring software of Agilent. Due to large sample variation between the groups, I am not getting significantly differential expressed genes after multiple test correction. What should I do now? Is the data is publishable without multiple test correction?

gene • 194 views
ADD COMMENTlink
0
Entering edit mode

I wouldn't recommend analysis without multiple testing correction. Try to find out why there is so much variation between samples of the same group.

ADD REPLYlink
0
Entering edit mode

How many biological replicates? What is the experimental design? Did you investigate the cause for the large variation between samples?

You may validate the genes of interest with real-time qPCR. If you are interested in the general picture, perform gene set enrichment analysis of metabolic pathways / GO categories, maybe you will have interesting results of biological significance. Also, a literature review supporting your findings may help.

ADD REPLYlink
0
Entering edit mode

Sir, I took PBMC as model for this study. I challenged the PBMC with virus. I took 2 biological replicates. I am not getting the exact reason for this variation. May be the individual genetics matters here.

ADD REPLYlink
0
Entering edit mode

Two replicates is too few, more so if the data is noisy. You can validate the genes of interest, or you can repeat the experiment to increase sample size.

ADD REPLYlink
0
Entering edit mode

As per Agilent Scientific person, 2 biological replicate in single colour array and 3 biological replicates in Two colour microarray are sufficient.

ADD REPLYlink
0
Entering edit mode

In your question you ask if this is publishable. Two biological replicates is absolute minimum for statistical tests but not recommended, and you can only get a way with it if you have very good reasons (for example if you only have two rare disease patients, but low budget or not enough time is usually not a good reason). FDR adjustment is mandatory for DE analysis with genomics data, if not used you are looking at false positives, so trying to publish results without these adjustments, would not be publishable (in good journals). If I am asked to review such studies, I would send it back to the editor for rejection without further review.

ADD REPLYlink
0
Entering edit mode

As per your own experience, and a considerable literature body, 2-3 replicates are not sufficient. Maybe you can use the results you have to estimate the sample size necessary to reach a given power, there are several packages available, e.g:

Sample Size and Power Calculation in Microarray Studies Using the sizepower package.

ADD REPLYlink

Login before adding your answer.

Similar Posts
Loading Similar Posts
Powered by the version 2.1