In genome wide association studies association is tested with whole of the genome so multiple testing is required. It has its cost in terms of false positive association and so the bar of significance is very high.

I would suggest Plink for analysis. It is free software and easy to use.

Good Luck.

Thanks Sir. What if I am working on just a small region of a candidate gene? Do I need to still adjust ?

The reason that we have to adjust the threshold for significance is that we are doing independent analyses and chances of type 1 error increase with each independent analysis. If the markers you are analyzing are close together then it would not be independent analysis so you may not need to adjust for the multiple testing. This is the theory. I can give you more specific advice if I know further details of your study and research question. Is it an association or linkage analysis you are performing? 

Dear Sir, I am doing a case-control association analysis and working on a single candidate gene. I have screened for all SNPs present in the coding regions of  this gene. In my study there are just 4 SNPs present. I have done independent logistic regression analysis for establishing association with disease status.All these markers are quite close but not that highly linked(the highest LD D' value being 0.55).  What do I do? stick with independent analysis or do multiple testing. And can the analysis done using SPSS. I really need your valuable advice.Thanks and Regards-Nuzhat.

Thank you providing this detail. As it is a candidate gene there would be prior evidence for its association with disease. If the SNPs are in LD with each other then I do not think you need to account for multiple testing. Usually in case control design odds ratio or relative risk is the statistic you compute. You can send me the data file and I'll run the analysis for you. Please send the file in text or excel.

My email is [email protected]

Ayub

Thank you for your suggestions and help Sir. I will surely write to you if need something. Just wanna know now, at what cutoff value of D' that two SNP markers can be said or are in linkage disequlibrium?

Best wishes and regards

Nuzhat.

D/ is a measure of correlation and cut off will be dependent on the purpose of the analysis. I suggest that you report values. Are you planning to write this study for publication or for thesis. What kind of regression did you perform. Why did you perform a regression analysis. Did you need other variables to include in the model. What kind of statistics you have for significance testing. 

I realize that you performed logistic regression. Please ignore my question about type of regression. Another issue you need to take in account is the population allele frequencies of the SNPs.

Dear Sir, I am working for both thesis and publication. I will keep in mind what all suggested.

Thanks and regards

Nuzhat 

Can you please share your work when it is published.

Sure

If your SNPs are in the same gene and/or in LD you may perform haplotype analyses rather than successive single SNP analyses that would require Bonferoni correction for multiple testing. For haplotypes analyses you may use the Thesias software or the PMplus EHplus software or the Haplotype Trend regression software.

Dr Vasseur

Your comment is illuminating. How would haplotype analysis affect the power?

Dr Vasseur

My SNPs are of the same gene and I have done haplotype analysis.  Prior to this I have also checked individual association by logistic regression analysis. Could you please elaborate what else is needed and if you could suggest some reference articles to follow. Thanks.

You can follow 2 papers I published that I think are available as full text:

Diabetologia 2005 vol. 48 (5) pp. 892-9

Hum Mol Genet 2002 vol. 11 (21) pp. 2607-14

Moreover you may follow papers using the Thesias software of David Tregouet