That doesn't look too bad but of course you have to do the last one almost 400000 times, so that will take a while. I haven't looked at your queries too closely but is there a reason why you need to issue a query for genotypes with a single sample and a single snp instead of querying for all samples IN (...) and all snps IN (...) in one go? That should be a lot faster. You can associated sample, genotype and snp in your perl script by parsing the results. Other than that I would still look at the indices, which might not be optimal for your queries, and also at the sorting you let the DB do - that too will take time and might not be necessary, depending on how you use the data afterwards.