Practical Issues in Genome-wide SNP Association Study

Abstract

The associations between SNPs and a disease in genome-wide case/control study are subject to a wide range of potential biases, so they may represent false positives. In the real world, the study designs frequently include selection bias, information bias, and confounding due to many practical difficulties. Thus, many non-replicating, non-causal associations are probably due to selection of non-comparable case/control subjects.

Many public and commercial SNP data analysis tools are available such as Plink, HelixTree and etc. It is worth noting that the analysis results could sometimes be quite different, because each tool may use different options or algorithms. Sometimes the different analysis results come from S/W bugs, especially in the free public analysis tools. Therefore, it is recommended to validate the analysis results with a second analysis tool and to verify the selected options before interpreting the study results.