Dear All, 

I am working on Micro-array data analysis  with this GEO acc no. GSE31747, In this paper PMID 22024983 (https://www.ncbi.nlm.nih.gov/pubmed/22028943) they have used ANOVA method to identify the significant genes,Now i am  trying to compare their results with Limma package but i am not getting any single significant genes(P-Value<0.05 ). So can any one tell me if Limma not give any significant genes what else other methods i can go.

By 

Mathavan M

Research Scholar Centre for Bioinformatics

Pondicherry University

mathavanbioinfo@gmail.com

Hi,

IMHO, the number of subjects is too small in order to get any nor general insights.

https://www.ncbi.nlm.nih.gov/sites/GDSbrowser?acc=GDS4356

When looking for differential genes concerning INFECTION vs MOCK, there is only one reported https://www.ncbi.nlm.nih.gov/geoprofiles?term=(%22infection%22[FINF])%20AND%20GDS4356[ACCN]. If INFECTION vs MOCK is your interest, limma gave you the right answer.

Best.

This data set clearly cannot be analyzed by a single factor ANOVA. It requires a two way approach with adjustment for donor affects. There is very little differential expression in this data apart from the large differences between the three donors. Nevertheless, when I analyzed this data in limma, I had no trouble with finding a few genes with P=2E-6 and FDR=0.004.

As far as I know, there is no valid statistical test with more sensitivity than the limma approach.

I would treat the reported results in the Wahl-Jensen et al article with a great deal of caution. It is not clear to me how they did the statistical analysis and there is no indication that they made any adjustment for multiple testing. Many or most of the DE results they report might not be reproducible.