Plot GWAS p value distribution for multiple traits
0
0
Entering edit mode
4.1 years ago

I have 10,000 SNPs and 48 different phenotypic traits, so I performed GWAS based on Mixed Linear Model (MLM) approach using Tassel. I have got associations and now I am interested to see the distribution of p values to better interpret p-values so I extracted the p-values for all traits from the association file to generate a matrix containing 10473 rows (row one contain header) and 49 columns (column one shows marker names and rest 48 shows p-values of different traits), sample input file is:

Marker  BM_day_HS1718   BM_day_HS1819   BM_HS1718
AX-94381147 0.30253 0.53712 0.37212
AX-94381170 0.00997 0.17313 0.00127
AX-94381338 0.21632 0.85283 0.26634
AX-94381448 0.65978 0.55148 0.78321

AX-94381449 0.34668 0.38499 0.67861

I have tried to do an extensive literature search, there are methods available to see the distribution of a single trait but I am interested to see for multiple traits, I found a method from scmamp library which is very close to my requirment but it first makes use of friedmanPost function to computes the raw p-values for the post hoc based on Friedman's test instead of plotting p-values I have in my matrix.

I have found another thread "How to represent multiple p-values" where they have recommended plotting effect size instead of p-values but my scenario is a bit different as I want to see the p-value distribution in order to decide the significance cutoff for GWAS results.

I have also tried to generate qq plot for multiple traits but it compared the expected vs actual p-values, I am not sure if it's good enough.

Any help/suggestion would be highly appreciated

GWAS Plotting R • 2.1k views
ADD COMMENT
0
Entering edit mode

why do you think a qq plot is not good enough? You can try a manhattan plot as well.

ADD REPLY
0
Entering edit mode

I have ploted the qq plot for all traits together but it depeicts expected vs actual p-values and the graph is not very much clear that's why i am trying to look for other options

ADD REPLY
0
Entering edit mode

For your case, I believe that the Manhattan plot will fit exactly what you need!

ADD REPLY
0
Entering edit mode

I have manhattan plots for each trait but is there any way by which I can plot manhattan plot for all traits together to see the p-value distribution

ADD REPLY
0
Entering edit mode

From the sound of it, you seems to be doing a phewas study. This package might be helpful https://pubmed.ncbi.nlm.nih.gov/24733291/

ADD REPLY

Login before adding your answer.

Traffic: 2894 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6