Computational protocol: Impact of the AHI1 Gene on the Vulnerability to Schizophrenia: A Case-Control Association Study

Similar protocols

Protocol publication

[…] Prior to any association analysis, it is important to make a priori calculation to know the statistical power of the study. This power varies attending to different parameters, such as the sample size, the risk allele frequency, the incidence of the disease and the risk (odds ratio) assumed for each polymorphism, among others. For this study, QUANTO software version 1.2.3 was used to calculate the statistical power to find associations between genetic polymorphisms and the risk for schizophrenia. We assumed an odds ratio (OR) between 1.3 and 2. We also set the incidence of schizophrenia at 1% and the inheritance model as additive. The estimations for our different samples are summarized in . Particularly, this power ranged between 0.149 and 0.999 depending on the sample size, the risk associated to each allele and the allele frequencies. Thus, for those polymorphisms with a low MAF, the statistical power to detect association with the disease is expected to be lower, especially if the risk is low. In our particular case, 21 of the 29 SNPs presented a MAF equal to or higher that 0.3, and only 3 SNPs had a MAF lower than 0.1. Therefore, we consider the statistical power of the study to be adequate.Genotypes were assessed for Hardy-Weinberg Equilibrium (HWE) in both patient and control samples by applying a χ2 test implemented with Haploview Program version 4 . As the deviation from HWE in the control sample could be indicating a genotyping error, those SNPs which deviated from HWE in the control sample (P<0.1) were discarded from the analysis.Differences in the allelic frequencies between patients and controls were evaluated with a χ2 test via Unphased program version 3.0.12 , . Moreover, the web-based application SNPStats was used to compare, by means of logistic regression, the genotypic frequencies between groups. Different inheritance models (codominant, dominant, recessive, and additive) were tested. ORs and 95% confidence intervals as indicators of the risk associated to each genotype were also calculated. Finally, if any SNP was found to be associated with the risk for schizophrenia, a linear regression analysis with SNPstats was used to evaluate the effect of those SNPs on several clinical scores (BPRS, PANSS and KGV). Bonferroni sequential test for multiple comparisons was applied to correct all the reported p-values.Regarding the haplotype analysis, haploblock frequencies were compared between patients and controls with Haploview Program version 4 . Moreover, frequencies of sliding windows of two, three and four-marker haplotypes were estimated through a retrospective likelihood algorithm and compared between patients and controls with the Unphased software version 3.0.12 , . It should be noted that haplotypes with a frequency below 0.03 were considered as rare and therefore discarded from the statistical analysis. A 1000-permutation run was performed in all haplotype analyses to better estimate the significance of the positive associations.We also performed a joint analysis by combining the affected individuals and healthy subjects from Spain and Germany. The linkage disequilibrium patterns between SNPs were determined to assess the validity of such type of analysis. Moreover, these analyses were only stratified according to the region of origin of each individual (Germany or Spain) since there were no differences in other possible stratification variables, for example ethnicity. PLINK software version 1.0.2 was used to apply two different CMH tests, which are based on an “average” OR that controls for the potential confounding due to the cluster variable. Two variants of the CMH tests were performed: a) a CMH test that performs the case-control association test but controlling the effect of the cluster variable (in our particular case, the cluster variable was the country of origin); b) a CMH test that is used as a measure of the heterogeneity between samples, which assesses whether a particular SNP varies between clusters. Finally, we also performed a haplotype analysis following the same procedures described above for the Spanish and the German sample separately. […]

Pipeline specifications

Software tools Haploview, snpStats, PLINK
Application GWAS
Organisms Homo sapiens
Diseases Psychoses, Substance-Induced