nominally significantp = 4.88x10 -4
Furthermore, we note that when comparing the signals we took for replication (based on case control analyses) with association results with BMI as a continuous trait derived from an independent set of samples from UKBB, there are more directionally consistent and nominally significant associations with BMI than expected by chance suggesting that amongst these loci, there may be additional real associations (binomial p = 4.88x10 -4 , and binomial p = 9.77x10 -3 , respectively, Methods , S9 Table ).” Despite the smaller sample size, the obese vs thin comparison had increased power to detect some loci ( S3 Fig ), including a recently discovered variant near FAM150B [ 32 ] (rs62107261, MAF = ~5%), which did not meet our p<10 −5 threshold to be taken forward for replication in obese vs controls analysis ( p = 2.36x10 -4 ).