This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Hypergeometric test ?

I have a set of micro array data set and have total number of genes 23233 and subset of two classes like A and B after DEG so i want to anaylsis these for MGI Do disease data set by Hypergeometric test so what i take total number of genes =23233 class A like 1801 class B like 1417 so what i did i used that particular disease count in total number of genes and class and used this formula population size P= 23233 number of success in population= 8 of that disease in (P=23233) and calssA is subset of P is S =1843 and sucess in class A is S =3 and hypergeometric test value is Probability of drawing 3 successes or more from a sample of 1843: P(X >= 3) = 0.020604834192780287. Probability of drawing exactly 3 successes from a sample of 1843: P(X = 3) = 0.018475183897383608

so my first question is is this is right method or values I am using second there most of the values less then 0.5 so they mostly are non signficant

rna-seq

1 answer

  1. No this is not the right method. A hypergeometric test is inappropriate for microarray data. There are a number of packages in bioconductor (limma, lumi, etc.) that implement more appropriate statistics. In general, a hypergeometric test is best suited for contingency tables of counts.
  2. 0.05, not 0.5 and make sure to correct for multiple testing beforehand.

I understand OP has already done the microarray analysis and they have identified two groups of genes of interest and want to know if they are enriched in genes involved in some disease. This leads to a contingency table. A hypergeometric test seems reasonable to me.

Thanks May I Ask what OP means ? As far as i understand we have done DEG ,GO and other analysis so we want to find out Gene enrichment for disease associated genes in each class or group and see any biasness?

What you're doing with the hypergeometric test looks OK to me.

Ah, I misread then!

Log in to answer this question.