Prediction and characterization of genetically regulated expression of asthma tissues from African-ancestry populations.

Abstract

Genetic control of gene expression in asthma-related tissues is not well characterized, particularly for African-ancestry populations, limiting advancement in our understanding of the increased prevalence and severity of asthma in these populations. We sought to create novel transcriptome prediction models for asthma tissues (nasal epithelium and CD4⁺ T cells) and apply them in a transcriptome-wide association study (TWAS) to discover candidate asthma genes. We developed and validated gene expression prediction databases for unstimulated CD4⁺ T cells and nasal epithelium using an elastic net framework. Combining these with existing prediction databases (N = 51), we performed a TWAS of 9284 individuals of African ancestry to identify tissue-specific and cross-tissue candidate genes for asthma. Novel databases for CD4⁺ T cells and nasal epithelial gene expression prediction contain 8,351 and 10,296 genes, respectively, including 4 asthma loci (SCGB1A1, MUC5AC, ZNF366, and LTC4S) not predictable with existing public databases. Prediction performance was comparable to existing databases and was most accurate for populations sharing ancestry with the training set (eg, African ancestry). From the TWAS, we identified 17 candidate causal asthma genes (adjusted P < .1), including genes with tissue-specific (IL33 in nasal epithelium) and cross-tissue (CCNC and FBXW7) effects. Expression of IL33, CCNC, and FBXW7 may affect asthma risk in African- ancestry populations by mediating inflammatory responses. The addition of CD4⁺ T cell and nasal epithelium prediction databases to the public sphere will improve ancestry representation and power to detect novel gene-trait associations from TWAS.