I'm conducting a meta-analysis on two genome wide association studies (GWAS), each consisting of 150 SNPs, for which I computed summary statistics for association. I'm using the R package meta and the function metagen. It seems to me that this function is able to combine studies but on single measurement effect (I have 150), one for each study.

For example, I may have 3 SNPs on each study

STUDY    SNP    OR    SE
GWAS_A  rs694739    0.6691  0.07588
GWAS_A  rs9858968   0.1091  0.01588
GWAS_A  rs1529267   0.9291  0.02588
GWAS_B  rs694739    0.6128  0.37344
GWAS_B  rs9858968   0.0332  0.27344
GWAS_B  rs1529267   0.3481  0.81284

so I would like an output in a table format (e.g., as in PLINK), with 3 rows and the corresponding pooling statistics. I know I could manage it with for loops or by grouping with the by function on the summary output of metagen, but I wonder whether there is another R package/function to easily accomplish this.

  • $\begingroup$ I have the same problem and I can't figure out a solution. I started by doing my meta-analysis in PLINK, but I want the confidence intervals of the meta-analysis, which PLINK does not give. I then tried metagen() function in the meta package, but I'm struggling to write a loop, because the output is a meta class object, and I don't know how to extract the results. I would be happy with any of these solutions: -A way of getting the CI from PLINK or -A way of extracting the results from the meta object to a table. Thanks! $\endgroup$
    – user52652
    Jul 23, 2014 at 16:18
  • $\begingroup$ See @Joana's comment as a question $\endgroup$ Jul 23, 2014 at 20:34

2 Answers 2


The MetABEL part of GenABEL does this. For 150 SNPs, you might find coding the loop yourself quicker than ensuring it's doing exactly what you want. (Neither should take very long)

  • 2
    $\begingroup$ GenABEL link is dead, it is removed from CRAN, too. $\endgroup$
    – zx8754
    Nov 10, 2021 at 14:51

A quicker option would be to use the built-in meta-analysis option from PLINK:

It is quite straightforward, and no additional coding is needed. You could then use the meta package from R to make forest plots of significant results. My experience with the meta package is that it isn't very practical when doing more than a few meta-analyses.


Not the answer you're looking for? Browse other questions tagged or ask your own question.