Our clumping algorithm is iterative and has been used previously10. We describe it here for the case of identifying lead SNPs among the set of SNPs reaching P < 5×10−8; the algorithm is the same when determining sets of approximately independent SNPs for other P value thresholds.
