Were the two datasets generated for the same exact individuals from the population? If yes, it would be possible to combine the datasets by augmenting the genotypes for each individual. If they were generated for different individuals, I would avoid combining them. Not only were they generated with different enzymes, but different sequencing platforms. Unless you can control for those in your GWAS, I wouldn't combine them.

I can't see a good way to make this work.

I would even be hesitant to combine SNPs from two identically prepared RAD-Seq libraries processed separately - I would always go back to raw data and reprocess as a group since the detection of alleles is dependent on the frequency of the alleles in the sample. Also, the ability to detect heterozygotes is dependent on sequencing depth per locus - which will be hugely different in your case.

This is the strange thing that many people don't get with RAD-Seq and it's variants, the data you get for one individual is dependent on the sequencing reads for that individual, the parameters used in processing, AND the other individuals processed at the same time. If you change any element in this, you will change the data. Your data output from the RAD-Seq processing pipeline is probabilistic rather than real, and you can't see which bits are accurate. Those changes created through any changes rapidly become significant causing population structure between differently processed samples.

I am certain if you combined these data and then ran a program looking for population structure you would find the two datasets separate. If you really want to try to combine them, as an academic exercise, have a try of this. A PCA would probably be sensitive enough to detect it - that would be my first check. If your data already has significant populations structure it may not be the first PC, but it will be there...

What I would do in your case is to perform two separate GWAS using the two datasets. If both identify a region of the genome, or the same SNP as relevant, then you've got some independent replication that that region is important. If you don't, then the sample size for each group may be too small - and you'll just need to increase your sample.

If you do combine the datasets for fun do post a PCA to let us see the results.

Article Scanning RNA-Seq and RAD-Seq approach to develop SNP markers...