Replication of index SNPs and fine-mapping of birth weight loci

Analyses for both the replication study and fine-mapping were performed using SNPTest (Marchini et al. 2007). We used linear regression to estimate beta parameters (β, change of birth weight in grams per copy of effect allele) and 95% confidence intervals (CI) of the association between genetic variants and birth weight. β parameters were adjusted for same covariates as in the admixture mapping scan. Within each birth weight locus we corrected for multiple testing using the simpleM method (Gao et al. 2008), which estimates the effective number of independent tests using a principal component analysis approach.