Aim and hypothesis: We propose a project that follows on from the recent GWAS paper from the adipogen consortium, and the related paper led by the GSK team that resequenced the coding regions of the ADIPOQ gene.

We would like to use ADIPOQ genotypes to test the hypothesis that adiponectin causally influences insulin resistance, type 2 diabetes and related traits. We propose a Mendelian randomization approach. The proposal will build on the efforts reported in the main GWAS paper.

We plan to:

1. Use several SNPs in the ADIPOQ locus. These include those that are in partial LD, but that together explain more of the variance in circulating adiponectin levels - as replicated in several studies analysed by the proposers (RISC, CoLaus, WTDundee). We believe that there are at least 3 independent signals at ADIPOQ, captured by 3 or 4 SNPs. These signals are tagged to varying extents in different studies. One SNP in particular is a low frequency, large effect variant that will almost certainly require direct genotyping in most studies. Depending on the GWAS chip used and imputation quality, different studies will likely need to genotype 1-4 SNPs. Some of the critical SNPs are on metabochip.

2. Limit the study to ADIPOQ SNPs. We estimate that the ADIPOQ SNPs explain greater than 3% of the variance in adiponectin levels. With enough studies we should be well powered to perform instrumental variables approaches on traits strongly correlated to adiponectin levels, such as fasting and M value based measures of insulin resistance, and lipids. This approach will have the advantage that we are only dealing with "cis" effects - SNPs most likely only acting on circulating adiponectin levels directly through the encoding gene. We will therefore avoid uncertainties due to pleiotropic effects.

3. Limit the Mendelian randomization tests to only studies with all three sides of the "triangles" in place - SNPs, adiponectin and the "secondary" outcome phenotypes. This approach will allow us to perform the more formal "instrumental variables" approach to mendelian randomization. We have extensive experience performing these analyses and will draft an analysis plan.

4. for all studies involved perform a meta-analysis of instrumental variables based estimates of the correlation between adiponectin and secondary traits.

Exposure variable: genotype at 4 SNPs (please see ablove).

Primary outcome variable: adiponectin

Secondary outcome variables: T2D statuse, BMI, HDL, triglyceride, waist hip ratio, fasting insulin, fasting glucose

Confounding variables: age, gender, center (if applicable).

Requirements for participation:

1. ability to genotype 1-4 specific SNPs in study (or subset of study that has adiponectin and secondary outcomes measured) - small possibility that the rare SNP will be imputable but not currently possible with Affy6.0 chip and latest 1000G data.

2. SNPs, adiponectin and main secondary outcome measured in at least 500 individuals.

3. Highly preferable - access to STATA - the statistical software package that can implement instrumental variables analyses most readily.