Integrating Genetic and Transcriptomic Data to Identify Genes Underlying Obesity Risk Loci

Abstract

Genome-wide association studies (GWAS) have identified numerous body mass index (BMI) loci. However, most underlying mechanisms from risk locus to BMI remain unknown. Leveraging omics data through integrative analyses could provide more comprehensive views of biological pathways on BMI. We analyzed genotype and blood gene expression data in up to 5,619 samples from the Framingham Heart Study (FHS). Using 3,992 single nucleotide polymorphisms (SNPs) at 97 BMI loci and 20,692 transcripts within 1 Mb, we performed separate association analyses of transcript with BMI and SNP with transcript (P_BMI and P_SNP, respectively) and then a correlated meta-analysis between the full summary data sets (P_META). We identified transcripts that met Bonferroni-corrected significance for each omic, were more significant in the correlated meta-analysis than each omic, and were at least nominally associated with BMI in FHS data. Among 308 significant SNP-transcript-BMI associations, we identified seven genes (NT5C2, GSTM3, SNAPC3, SPNS1, TMEM245, YPEL3, and ZNF646) in five association regions. Using an independent sample of blood gene expression data, we validated results for SNAPC3 and YPEL3. We tested for generalization of these associations in hypothalamus, nucleus accumbens, and liver and observed significant (P_META<0.05 & P_META<P_SNP & P_META<P_BMI) results for YPEL3 in nucleus accumbens and NT5C2, SNAPC3, TMEM245, YPEL3, and ZNF646 in liver. The identified genes help link the genetic variation at obesity risk loci to biological mechanisms and health outcomes, thus translating GWAS findings to function.

Figure 1.. General workflow of the study design.

A) Step 1 included single omics associations for SNP to gene expression (P_SNP) and gene expression to BMI (P_BMI). B) Step 2 included the correlated meta-analysis to account for the interdependence between P_SNP and P_BMI. C) Identifying all SNP – Gene – BMI combinations that met our filtering criteria, which included correlated meta-analysis results that are more significant than individual omics associations. D) All significant SNP – Gene – BMI combinations were followed by validation in blood, liver, and brain tissues.

Figure 2.. Regional association plot

including association results for the discovery sample (Framingham Heart Study) for SNP with gene expression (blue), gene expression with BMI (green), and the correlated meta-analysis for SNP ~ gene expression ~ BMI (red). Annotation for potential candidate cis-regulatory elements from ENCODE are included for each reported SNP in the region. A. SNAPC3, B. YPEL3.

Figure 3.. Summary of validation and generalization for most significant SNP in discovery corelated meta-analysis.

Results are provided for discovery sample (FHS, blue), validation in blood (CCHC, red), and generalization to hypothalamus (Hypo, green), nucleus accumbens (Accum, purple), and liver (yellow) tissues. We provide individual effect estimates and P-values for each ‘omic and meta-analysis. Filled diamonds indicate significant associations in the meta-analysis (Note: FHS is noted as NULL, as all are significant).