Kinship coefficients and F ST , which measure genetic relatedness and the overall population structure, respectively, have important biomedical applications. However, existing estimators are only accurate under restrictive conditions that most natural population structures do not satisfy. We recently derived new kinship and F ST estimators for arbitrary population structures [1,2]. Our estimates on human datasets reveal a complex population structure driven by founder effects due to dispersal from Africa and admixture. Notably, our new approach estimates larger F ST values of 26% for native worldwide human populations and 23% for admixed Hispanic individuals, whereas the existing approach estimates 9.8% and 2.6%, respectively. While previous work correctly measured F ST between subpopulation pairs, our generalized F ST measures genetic distances among all individuals and their most recent common ancestor (MRCA) population, revealing that genetic differentiation is greater than previously appreciated. This analysis demonstrates that estimating kinship and F ST under more realistic assumptions is important for modern population genetic analysis.Kinship coefficients and F ST are defined as probabilities of identity-by-descent [3][4][5]. Kinship matrices are crucial for accurate inference under population structure in many important biomedical applications, including genome-wide association studies [6][7][8][9][10][11][12][13] and heritability estimation [14,15]. However, the most commonly-used standard kinship estimator [9,10,[13][14][15][16][17][18][19] is accurate only in the absence of population structure [2,20]. Likewise, current F ST estimators assume that individuals are partitioned into statistically-independent subpopulations [4,5,[21][22][23], which does not hold for human and other complex population structures. The human genetic population structure is remarkably complex, shaped by geography and population bottlenecks in migrations out of Africa [24][25][26][27][28][29][30][31][32][33][34] and admixture events [35][36][37][38][39]. We use human data to illustrate the improvements provided by our new approach.Models and methods. Our new kinship and F ST estimators were derived assuming arbitrary population structures, and they yield nearly unbiased estimates [2]. Suppose there are n individuals Independent subpop. model