In silico analysis
In silico analysis was performed to predict the effect of ANGPTL8 p.Q121X (rs145464906) mutation on transcription. We predicted whether the ANGPTL8 p.Q121X mutation resulted in nonsense mediated decay using RNA sequencing transcript isoform data publically available from the Genotype-Tissue Expression (GTEx) project (http://www.gtexportal.org/) and a predictive model implemented in MAMBA (http://www.well.ox.ac.uk/~rivas/mamba) [7, 8].
Analysis was limited to subjects of self-reported European ancestry. Studies contributing summary results included the BioImage Study , the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium Diabetes Working Group , and studies from the T2D-GENES consortium with > 5 minor alleles for ANGPTL8 p.Q121X (KORA and UK) . The BioImage Study and the KORA and UK studies from the T2D-GENES consortium have been approved by the MIT IRB (#1010004095, #0912003615 #1107004579, respectively). The CHARGE Diabetes Working Group consists of 27 European Ancestry studies and all participating studies were approved by local institutional review committees. All subjects have provided written informed consent.
Genotyping and quality control
All study participants were genotyped on the HumanExome BeadChip v.1.0 or v1.1 (Illumina) for ANGPTL8 p.Q121X and called using joint calling , GenomeStudio or zCall . Quality control involved checking concordance to GWAS data and excluding those individuals missing >5 % genotypes, population clustering outliers, individuals with high inbreeding coefficients or heterozygote rates, individuals with gender mismatches, one individual from duplicate pairs, and individuals with an unexpectedly high proportion of identity-by-descent sharing, with consideration for family studies, based on high-quality variants. All contributing studies used an additive coding of variants to the minor allele.
Association analyses were performed for fasting glucose and type 2 diabetes individually by cohort and summary statistics were shared. In the BioImage Study, type 2 diabetes was defined as individuals taking a medication for diabetes, having a fasting glucose > 126 mg/dl or having been told that he/she had diabetes. Type 2 diabetes was defined in the CHARGE Consortium Diabetes Working Group according to Wessel et al , and in the T2D-GENES Consortium according to Voight et al . Individuals with a diagnosis of diabetes were excluded from the fasting glucose and insulin analyses to avoid the variable effects of diabetes medications on the quantitative traits.
Two primary analyses were performed separately by study in available participants: (1). linear regression of fasting glucose levels with an additive coding of ANGPTL8 p.Q121X adjusting for age, sex, and principal components of ancestry; and (2). logistic regression of type 2 diabetes with an additive coding of ANGPTL8 p.Q121X adjusting for age, sex, and principal components of ancestry. Counts of the number of ANGPTL8 p.Q121X mutation carriers were obtained in type 2 diabetes cases and controls. Additionally, in 47,388 we analyzed the association of ANGPTL8 p.Q121X with fasting insulin levels adjusted for BMI, age, sex, and principal components of ancestry.
For the outcomes of fasting glucose and fasting insulin levels, statistical evidence from each study was combined using fixed-effects inverse variance meta-analysis. Heterogeneity of effects between studies was ruled out (p > 0.05).
For the outcome of type 2 diabetes, we performed meta-analyses by using Cochran-Mantel-Haenszel statistics for stratified 2x2 tables. The Cochran-Mantel-Hanszel method combines score statistics rather than Wald statistics. As an alternate approach, we performed a fixed-effects inverse variance meta-analysis on the betas and standard errors from the individual study results. Using a fixed-effects inverse variance meta-analysis did not change the results.
All analyses were performed using the software program R version 2.15. We estimated statistical power to detect association of ANGPTL8 p.Q121X with fasting glucose and type 2 diabetes using the Genetic Power Calculator (http://pngu.mgh.harvard.edu/~purcell/gpc/) .