Question: Fold change > 2 and P-value<0.05 in micro-array differential analysis
2
gravatar for Shicheng Guo
4.2 years ago by
Shicheng Guo7.8k
Shicheng Guo7.8k wrote:

Hi colleagues, 

I saw lots of high-through microarray or RNA-seq analysis, such as mRNA microarray, RNA-seq, use fold change > 2 and P-value <0.05 as the threshold of differential expression or differential change? How does it come? Will be acceptable without multiple-test correction? 

Best regards, 

fold change microarray pvalue • 4.0k views
ADD COMMENTlink modified 4.2 years ago by seidel6.8k • written 4.2 years ago by Shicheng Guo7.8k
2
gravatar for seidel
4.2 years ago by
seidel6.8k
United States
seidel6.8k wrote:

These are first pass minimal rules of thumb to evaluate a data set. The actual criteria applied would likely be data set dependent. Without multiple test correction it would be hard to publish this as a primary result, but if you used this criteria to select genes of interest and that led you down a path for follow-up experiments, then this result is not the center piece and you can get away with a maximal un-adjusted p-value. But again, this is usually and example of the least stringent criteria. If you can adjust the p-values (and still see significance), then you should. Likewise, some data sets will have significant genes by p-value yet smaller fold-change values (i.e. 1.4 fold). The criteria you chose are typically a function of how the results will be further evaluated, and how stringent you can be without losing everything. If you adjust the p-values and all significance goes away - it doesn't necessarily mean your experiment failed (but it might), but it does suggest you will have lots of false positives. In the other direction, if you adjust p-values and apply increased stringency (e.g. adj. palue < 0.0001 and FC > 1.5), and you still end up 3000 genes, when your downstream experiments can only deal with 100 genes - then you would increase your stringency even more. In summary - the choices you make depend on the characteristics of the data set. The criteria you listed are simply a very minimal stringency.

ADD COMMENTlink written 4.2 years ago by seidel6.8k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1267 users visited in the last hour