(b) Potential issues for performing a meta-analysis across
studies: one is that different genotyping platforms may
be used, and different analysis strategies might have been
used in the beginning, and the definition of cases and
controls may differ as well.4 Notes
- Two platforms offer different approaches to assay SNP varia-
tion. Illumina uses bead-based technology to recognize SNP
alleles, while Affymetrix uses DNA sequence as a spot on the
chip to detect a specific SNP allele [15]. Take into consider-
ation that different SNPs should be selected by the platform for
different study aim of human population. For example, more
SNPs are required with higher genomic coverage for African
population study, because African genomes have been existed a
longer time and therefore have less LD between alleles at
different SNPs. The genomic coverage in Africans was lower
due to greater genetic diversity and weaker LD. - Vague or nonexistent consensus definition may increase etio-
logical heterogeneity of a disease. An evidence-based and well-
defined phenotype is the best used for a genome-wide associa-
tion study. - If the population-based heritability of common disease is lower
than 20%, a large sample size is needed to identify the propen-
sity of genetic variants and hence increase the difficulty of the
study. - Well-characterized samples have long been used in the golden
rule of epidemiological study design for case-control analysis.
Population stratification refers to the most important type of
bias that may produce false-positive association results in case-
control studies. It can be minimized if the controls are matched
to cases when cases and controls are restricted to a particular
ethnic group. - Large sample size is one of the determinant factor for a suc-
cessful GWAS that ensure the genuine associations rank at the
top of the SNPs according to theP-values [16]. - You can also obtain genotyped controls and cases from data-
base, such as the Wellcome Trust Case Control Consortium or
UK Biobank, which is a more efficient way but may lack data
that you need in your study. - For a standard two-stage study design, genotyping a large scale
of markers on available samples and usingP-value threshold to
identify potential signals of association is the end of the first
stage. In stage two, we usually genotype the top signals on the
An Overview of Genome-Wide Association Studies 105