There was strong deviation from the uniform distribution of P-values expected under the null hypothesis (Supplementary Fig. 1). This deviation was attenuated, but persisted, following removal of those signals associated with the two previously identified loci. In total, 945 SNPs representing 45 loci (r2 <0.05 based on HapMap in a 750 kb region) were associated with age at menarche at genome-wide significance levels (P<5×10−8) (Fig. 1; Supplementary Fig. 2). None of these were located on the X chromosome. These 45 loci included three apparent second signals (defined as two genome-wide significant SNPs in low LD (r2<0.05) in the same 750 kb region) at 2q33.1, 6q21 and 14q32.2. The second signal at 6q21 (rs314279) had a low minor allele frequency (MAF, 6%) and was not present in many studies. We therefore genotyped this SNP de novo in the InChianti cohort and found it was in LD with the top chromosome 6 signal (rs7759938, r2=0.3). In HapMap the r2 between the two chr 6 SNPs was 0.015, but the D’ was 1.0. To verify the independency of additional loci, we performed a
