An X Chromosome Transcriptome Wide Association Study Implicates ARMCX6 in Alzheimer’s Disease

Figures

• 

  Figure 1.. The working flow of model training using GTEx data. Elastic net regression models of the X chromosome gene expressions were trained in whole blood and brain cortex with different mixing parameters, in both sex-stratified and sex-combined manner. Genetic variants on the X chromosome were coded as 0/1/2 in female samples and 0/1 in male samples. Model comparisons with two-way ANOVA

• 

  Figure 2.. Comparison matrix of the balanced sex-stratified elastic net models trained in GTEx whole blood and brain cortex data (α = 0.5). The four diagonal panels are the distributions of model R2. The number of genes being modeled in each tissue-sex set is listed in the distribution plot. The off-diagonal panels compared the R2 of the models in pairs of tissue-sex sets. The four rows are for the female-blood, female-cortex, male-blood, and male-cortex. For each row, the genes of the significant models in the tissue-sex set were matched in the other sets for the matching R2. The diamonds represent genes with a significant model (R < 0.1 and p > 0.05) in another tissue-sex set, while the circles represent genes with a non-significant model in another tissue-sex set. The number of gene models and the Pearson correlation coefficients were listed in the comparison plots.

• 

  Figure 3.. The number of SNPs in the validation datasets and the number of genes predicted. A and B: Venn diagrams representing the number of SNPs in the tissue-matching validation datasets and our sex-stratified prediction models. C and D: Venn diagrams illustrating the number of genes predicted by the sex of the models and validation samples.

• 

  Figure 4.. Genetic architecture of TWAS significant hit, ARMCX6. All three panels show the same region. A: The positions and effect sizes of the 23 selected SNPs in the cortex-specific expression prediction model trained with female-only samples. B: Hudson plot of eQTLs of ARMCX6 in female-only GTEx cortex samples. C: Hudson plot of ADGC meta-analysis GWAS in female-only samples. The red dots in B and C represent the SNPs in the expression prediction model.