Understanding Conditional analysis

Question

[OborneE]

Hello,

I'm new to data analysis and I'm having some trouble understanding the process and need help with a few questions.

I am running a conditional analysis on chromosome 19 in relation to Alzheimer's disease, I understand that conditional analysis tests whether SNPs have association independent of the SNP/SNPs you're conditioning for. I want to test if the SNPs in chr 19 are independent of my top SNP rs2075650.

I have run the conditional analysis on ~9500 SNPs from chr 19. I have read that you continue to add SNPs from the results to the list of SNPs you are conditioning for and re-running the analysis until the results are no longer genome wide significant.

Are the SNPs you add to the list secondary signals?

So far I have done the above 33 times and still the large majority of results from the results file are genome wide significant, do I just carry on? as it seems like there will be many hundreds more to add.

I have also read elsewhere that once you condition for your top SNP that SNPs remaining significant are assumed to have an independent effect on the phenotype with respect to the conditioned SNP. Is this true, otherwise why do the above?

About 200 of the 9500 SNPs are no longer significant after conditioning for my top SNP, does this mean those 200 are not independent of my top SNP?

Thanks, Emma

[alessandrotestori7]

Hi Emma, I am also doing a conditional analysis in order to see how many independent significant SNPs are in my locus of interest. Would you advise to condition all the SNPs in the locus on the lead SNP (SNP with lowest p-value) recursively, till there is no more significant SNP left? Do you remember where you have read this? Please let me know. Thanks, Alessandro.

Answer 1

[Veera]

Conditional analysis to identify the independently associated SNPs, can be done just with the summary statistics using GCTA software. (link: http://cnsgenomics.com/software/gcta/cojo.html)

[lovegenetics]

Hi Veera, the link/page is not found. Can you share the recent/up to date link please?

[GenoMax]

https://cnsgenomics.com/software/gcta/#Overview

[moxu]

I read the documentation but still feel confused about how conditional analysis is done with only summary stats (no individual geno/pheno info). My understanding is that when you do conditional analysis, you take a stepwise like procedure to add more snps to you linear regression (e.g. bmi ~ top_snp + candidate_snp), and this needs individual level data.

Could you please enlighten me a little bit?

Thanks so much!