Rare protein-coding variation and the genetic architecture of height in >1.4 million individuals

Highly heritable, polygenic, and easily measured, adult height has long been the model trait in human genetics. While the landscape of height-associated common genetic variation has been studied extensively, rare variation remains relatively unexplored. Using rare protein-altering variants in a discovery set of 826,066 exomes, we identify 207 height-associated genes - 98% of which replicate in an additional 624,567 individuals. The rarest and most deleterious class of variation, singleton (frequency <0.0001%) putative loss-of-function (pLoF) variants implicated 17 genes with large effects on height ranging from -17 cm (ACAN) to +11 cm (FBN1) per allele, 52x larger than the average effect of common height-associated variants and comparable to the 1% tails of a common variant polygenic score. Several genes (e.g., TET1, DTL, IGF2BP2) have effect sizes at least as large as established Mendelian height genes but lack documented stature or skeletal growth syndromes. This is particularly true for genes in which rare variants associate with increased height. We performed the largest rare-variant study of height to date, directly implicate 207 genes that broadly overlap with both GWAS associations and Mendelian height syndromes, assess the impact of rare variants on heritability and prediction, provide evidence that height is an underappreciated clinical feature of Mendelian disorders, and demonstrate the utility of large population-scale sequencing studies for classifying individual variants and dissecting complex trait architecture.