SNP studies have become a popular tool to study diseases. Recently, there were several studies that explored more than 8-12 SNPs in one disease, several of them were from big centers and published in high impact factor journals, using p
Yes, technically you are correct, but you are forgetting about the supposed independence or relationships between SNPs. If there are lets say, 8 SNPs that are independent from each other (so not in any LD), each test is being considered against a P value of 0.05 because they are independent tests. If these were, say 8 SNPs all from a single gene and in a single LD block, then yes, you would want to use a method such as Bonferroni, where you would use something more like 0.05/8 (=0.00625) as your cut off for "significance".
If those SNPs are in linkage disequilibrium then they should not be considered independent, and would require some multiple correction to the P value. However if they are not in linkage disequilibrium at all, then yes there is an argument for the idea that this would still apply.
If SNP are in LD, they are not independent and they should be treated as if they were less than the genotyped number. However, even if they are independent, you shouldn't use the 95% threshold. If you use this threshold and you carry out 100 experiments, you will find 5 that just by chance follow your hypothesis. A simple, yet drastic solution is dividing the threshold by the number of SNP you consider (this is the Bonferroni criterion). A more refined method is the False Discovery Rate (http://en.wikipedia.org/wiki/False_discovery_rate). The 95% was developed by the "frequentist" school of statistics, which is strongly criticized by the "bayesian" school. Just to cite my friend Agustin Blasco, if you have the chance of 5% to be killed today in a car accident, you most probably don't get you car, but if you have a 80% probability to win some money on a game, you'll bet! So don't take the 95% too seriously, unless your referee is a frequentist :)
Hi Alessio: Thank for lighting up my mind. Your comment is important. I guess when we submit a manuscript we have to avoid one group of biostatisticians.
Hi Ru-Jeng, it's difficult to know in advance to which school of thinking a referee belongs, but for sure you can use the above arguments to defend your valid results against stupid statements like "you only reach 94% of confidence so I reject your manuscript". cheers, alessio
Dear Alessio: Unfortunately once the reviewers reject the manuscript there is no way we can argue. That is the truth and essence of "unbiased peer review". Ciao. Ru