[R] Manhattan Plots for GWA studies

Will will.eagle at gmx.net
Mon Nov 24 13:24:24 CET 2008


I want to create manhattan plots for genome-wide association studies 
(-log(p) ~ SNPs).  I wonder whether there is a recommended approach 
avoiding the problems I have with the functions named below

Im my data I have only significant SNPs for dozens of similar phenotypes 
using different association models. This means some chromosomes are not 
represented and some SNPs multiple times in my data, which could cause 
some confusion in the plotting functions below. This is what my data 
looks like:

	chrom 	pos 	p 	trait 	model
SNP1	1	10	1E-05	qt1	simple	
SNP2	1	20	1E-10	qt2	simple
SNP2	1	20	1E-11	qt3	adjusted
SNP3	1	40	1E-02	qt4	simple

Solutions so far:

The simple *plot ()* function prints error bars in my data as some SNPs 
are represented multiple times (due to multiple phenotypes/models). 
*sunflowerplot()* works rather well. Both functions label the x-axis 
with snp names which are not readable if a great number of SNPs examined 
at once.

I have found the *mhtplot ()* function in the <gap> package. However, it 
gives error messages and just plots fragments of the plot.

I found the *wgplot ()* function 
(http://bioinfo-mite.crb.wsu.edu/Rcode/wgplot.R)  which tries to fix 
some bugs in mhtplot(). This works rather well, but it sometimes screws 
up with the x-axis labels which are printed on the x-axis or are 
sometimes missing.

Any suggestions?

Thanks in advance,


More information about the R-help mailing list