Study Design and Methods

Characteristics of cases and controls

The cases and controls came from two sources. The 488 cases and 498 controls successfully typed from the Jackson Heart Study were all extensively phenotyped individuals, collected from a larger cohort of 5,302 African Americans from Jackson Mississippi who are being studied for phenotypes related to heart disease . The cases were ascertained based on fasting serum glucose > 125 mg/dl, use of medications for diabetes, or positive history of diabetes plus hemoglobin A1c > 7. The controls were ascertained based on negative personal history, fasting blood glucose <100mg/dL, hemoglobin A1C < 6, and not being on medications for T2D. The controls were further chosen to be matched to cases in terms of their proportion of females (66%), and to have high body mass indexes (average = 31.4) and older age (average = 60.1) to make it more surprising that they did not have T2D.

The 990 cases successfully typed from the Multi-Ethnic Cohort of Los Angeles and Hawaii (MEC) came from a study of 215,251 individuals from Los Angeles and Hawaii, who have been followed by questionnaire since 1993-1995. These individuals are primarily being studied for cancer-related phenotypes; however their questionnaires also reported information about diabetes status, and medications. The samples included in this study are 65% female, all reported having Type 2 Diabetes, and also reported using medications specific to T2D.

  Cases Controls
Jackson Heart Study 488 498
Multi-Ethnic Cohort Study 990 0

Genotyping and data curation

We used the Illumina BeadLab platformi at the Broad Institute of Harvard and MIT to successfully genotype 1,291 markers chosen to be extremely different in frequency between west Africans and European Americansii,iii. Data filtering and cleaning procedures were identical to those reported in a recent study on admixture mapping of multiple sclerosis genesiii. Data for these markers (inferred frequencies in west Africans and European Americans, and scores for association) are reported in the attached table [link to table here].

Data analysis

We used the ANCESTRYMAP softwareiv to analyze the data, using the same strategy that was reported in a recent study on admixture mapping of multiple sclerosis genesiii.

We carried out the analysis assuming a prior distribution for ancestry risk that tested both for loci associated with increasing risk due to European ancestry, and increasing risk due to African ancestry. Thus, the prior distribution considered equally likely models of 0.6, 0.7, 0.8, 0.9, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, and 1.8-fold increased risk for European ancestry, and simultaneously tested for the opposite risk in controls at the same locus.

The ANCESTRYMAP program calculates a log-factor for association at equally spaced points on the genomeiv. A local score of 5, for example, means that the data at that locus are 105=100,000 times more likely under one of the disease models than under the null model. The score must cross a high threshold of > 5 to be considered genome-wide significant (>4 to be suggestive), since there are many loci in the scan.

ANCESTRYMAP also calculates a genome-wide score for association, averaging the Bayes factors over equally spaced points, to assess evidence for association anywhere. A score of >2 is considered genome-wide significant (>1 to be suggestive).

A Note about early access to the data

We are making the genome scan results publicly available prior to publication so that researchers interested in the genetics of T2D can use the results of the scan to prioritize follow-up of any regions of interest. We plan to publish the results of the scan, with much more detailed discussion, in due time.