Question: microarray data analysis
0
gravatar for au.rinki.bio
23 months ago by
au.rinki.bio10
University of Allahabad, India
au.rinki.bio10 wrote:

I am working on microarray data analysis. In Bonferroni correction α is devided by total sample number of microarray experiment or gene number.

multiple testing • 728 views
ADD COMMENTlink modified 23 months ago by WouterDeCoster42k • written 23 months ago by au.rinki.bio10
1

I'd strongly urge you to use Benjamini-Hochberg, if it is available to you, rather than the unforgiving Bonferroni approach. Strictly, in Bonferroni, alpha should be divided by the number of hypotheses tested, which is typically (the number of contrasts X the number of features), so in your terms, if you're only testing a single contrast it would be the number of genes

ADD REPLYlink written 23 months ago by russhh4.8k

thank you so much for giving suggestion. actually i am new in this field i have one more question i am working on microarray data, i have a GSE14007 for the analysis. i have normalyzed this data and tried to find out differentially expressed genes. GSE14007 has 12 sample for normal condition and 12 sample for treated condition. after the preprocessing of this data their arround 22000 genes. to find out DE gene i have apply t-statistic and filtered the gene which p-vale<0.05. from litrature review i have found p-adjusted value and fdr based selection are more appropriate. now first i want to go with benferoni correction,alpha should be divided by the number of hypotheses tested but i am confuse here what is tested hypothesis? in my case is total sample number 24? or total gene number 22000? i hope you understand much better now. kindly give me your suggestion.

ADD REPLYlink written 23 months ago by au.rinki.bio10

I understood you to begin with, which is why I stated that you had to divide alpha by the "number of hypotheses tested". You only have one contrast (treated vs normal), and because this defines a hypothesis for each of your genes, you should divide by the number of genes.

But please be aware that the Bonferroni correction is too stringent, and is rarely used in microarray analysis. You should use the BH correction that I mentioned.

ADD REPLYlink written 23 months ago by russhh4.8k

Thank you so much for insightful reply.

ADD REPLYlink written 23 months ago by au.rinki.bio10

What is your question?

ADD REPLYlink written 23 months ago by Michael Dondrup46k

I interpreted this:

In Bonferroni correction α is divided by a) total sample number of microarray experiment or b) gene number?

ADD REPLYlink written 23 months ago by WouterDeCoster42k

i am working on microarray data, i have a GSE14007 for the analysis. i have normalyzed this data and tried to find out differentially expressed genes. GSE14007 has 12 sample for normal condition and 12 sample for treated condition. after the preprocessing of this data their arround 22000 genes. to find out DE gene i have apply t-statistic and filtered the gene which p-vale<0.05. from litrature review i have found p-adjusted value and fdr based selection are more appropriate. now first i want to go with benferoni correction,alpha should be divided by the number of hypotheses tested but i am confuse here what is tested hypothesis? in my case is total sample number 24? or total gene number 22000? i hope you understand much better now. kindly give me your suggestion.

ADD REPLYlink written 23 months ago by au.rinki.bio10
0
gravatar for WouterDeCoster
23 months ago by
Belgium
WouterDeCoster42k wrote:

Multiple testing correction is performed with regard to the number of statistical tests you are doing. Sinc differential expression analysis tests per gene whether this gene is differentially expressed, the number of tests is the number of genes.

ADD COMMENTlink written 23 months ago by WouterDeCoster42k
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: 1841 users visited in the last hour