Computational protocol: Gene-based analysis of genes related to neurotrophic pathway suggests association of BDNF and VEGFA with antidepressant treatment-response in depressed patients

Similar protocols

Protocol publication

[…] Imputation was carried out using IMPUTE2 v3, with haplotype reference panels released in March/April 2012 from the 1000 Genomes Project on the basis of HapMap build 37 (https://mathgen.stats.ox.ac.uk/impute/data_download_1000G_phase1_integrated_SHAPEIT2.html). Only imputed SNPs with high genotype information content (i.e. IMPUTE info score > 0.5) were used in association analyses. In total, 30,040,257 SNPs were imputed with high confidence for each individual in the samples. Following the same quality control procedures for markers, a total of 4,241,701 SNPs were retained for analyses. Gene-mapping was conducted using 50 kb upstream and downstream of the gene boundaries. [...] We performed single-marker association and gene-based association tests for treatment-response phenotypes. Both linear and logistic regression analysis were conducted with additive genetic model. All models were adjusted for gender and age to correct for different distributions in gender and age across treatment-response phenotypes in MDD patients.Gene-based association analyses were conducted to obtain gene-level empirically significant estimations. Information from a set of SNPs (association p-value < 0.1 by default) within a gene was aggregated. To account for the linkage disequilibrium (LD) among markers, only SNPs having r2 < 0.5 with each other were retained for each gene. We also fitted both linear and logistic regression models to perform gene-based association tests. Empirical p-values were calculated with 50,000 permutations.To combine individual association results (i.e. p-values) from the two samples (i.e., NHRI and TVGH), meta-analysis was applied using the inverse Gamma model with a shape parameter (α) of 1, that is, the Fisher’s method to summarize association information across the two samples. Alternatively, we conducted association analyses to an integrated database of NHRI and TVGH. We adopted a method of LD adjusted multiple testing correction developed by Duggal et al. to account for the interdependence among SNPs to balance between false-negative and false-positive findings. Only SNPs and genes with p-values less than 5.0 × 10−4 (or 5.0 × 10−3) and 5.0 × 10−2 (or 1.0 × 10−2), either in mega-analysis or in meta-analysis, were considered to be significant (or suggestive), and were reported in single-marker association analysis and gene-based association analysis, respectively. Additionally, we conducted a case-control (455 MDD patients and 2,998 healthy controls) study using SNP-based and gene-based association analyses to investigate whether the treatment-response associated loci were also MDD susceptible loci.All aforementioned analyses were conducted with R version 3.0.2, PLINK version 1. 90b3.37 64-bit, and haploview version 4.1. Additionally, to explore the potential roles of these six SNPs as expression quantitative trait locus (eQTL), we used HaploReg (http://compbio.mit.edu/HaploReg) to search gene regulation databases. […]

Pipeline specifications

Software tools IMPUTE, SHAPEIT, PLINK, Haploview, HaploReg
Application GWAS
Organisms Homo sapiens