Replication in an independent cohort is of course the gold standard in GWAS studies, and many high profile journals will now (quite rightly) not accept finding indicating a phenotype genotype association without evidence of replication.
However GWAS is no longer the domain only of human genetics and is being increasingly used on microbial species, pathogenic on humans, animals, and plants. However having read many of these studies (mostly in plant pathology) i can find no discussion of technical replication practices in their methods. In my own research I intend to conduct GWA between snps and gene presence-absence genotypes gathered by illumina sequencing, and phenotypes of virulence (as a continuous quantitative variable) when inoculated onto the host plant (which is wheat), of a large population of a fungal pathogen. i have inoculated 120 plants with 120 isolates of the pathogen, and i'm an additional replication cohort of 80 plants and isolates.
My question is, what is required to make these additional inoculations an 'independent' cohort?
The idea of independence in technical replicates is to correct for systematic bias present in the discovery cohort, I will likely grow these plants in a separate greenhouse bay for example (although since the different host-isolate combinations are randomised within the greenhouse even this may not be necessary); in a human GWAS different genotyping methods would be ideal in the replication cohort due to possible introduction of bias in the snp-chip, in my own study i'm unsure if this would be necessary as variants in my data will undergo stringent filtering prior to genotype calling.
If anyone could help point out to me what my possible sources of systematic biases are, i'd be very grateful.
The concept of independent replication has to do with how genealisable you want your results to be. For example, you have inoculated 120 plants with 120 isolates of the pathogen. If you are only interested in making inferences about these 120 plants, then you do not need another sample of plants; but if you want to make inferences about plants in general, then your replication sample must consist of 120 different plants. Similarly, if you are only interested in these 120 isolates of the pathogen, then you do not need another sample of isolates; but if you want to make inferences about isolates of this pathogen in general, then your replication sample must consist of 120 different isolates. You should not look to the literature, but rather to your specific purpose, to decide what constitutes an independent sample.