Analysis of the genetic basis of height in large Jewish nuclear families

Abstract

Despite intensive study, most of the specific genetic factors that contribute to variation in human height remain undiscovered. We conducted a family-based linkage study of height in a unique cohort of very large nuclear families from a founder (Jewish) population. This design allowed for increased power to detect linkage, compared to previous family-based studies. Loci we identified in discovery families could explain an estimated lower bound of 6% of the variance in height in validation families. We showed that these loci are not tagging known common variants associated with height. Rather, we suggest that the observed signals arise from variants with large effects that are rare globally but elevated in frequency in the Jewish population.

Fig 1. Founder effect in Ashkenazi Jews.

(a) Minor allele frequency of variants in 7509 Europeans (X-axis) vs. their allele frequency in 151 Ashkenazi Jews (Y-axis). Each dot represents one of ~90.5 million genetic variants from the gnomAD database. Color displays density, i.e. the number of variants in that region of the plot, on a log10 scale. The magenta rectangle encloses all variants that are rare (MAF<1%) in Europeans but common (MAF≥5%) in Ashkenazi Jews. Black line shows y = x. (b) Same as in (a), but with allele frequency in a random sub-sample of 151 Europeans shown on the Y-axis.

Fig 2. QTL mapping.

(a) For each LOD detection threshold, the number of detected QTLs is plotted in blue, the expected number of false-positive QTLs based on permutation analysis is plotted in red, and the difference between these two numbers is plotted in green. (b) Permutation-based false discovery rate at each detection threshold (c) Statistical significance of the number of detected QTLs at each threshold, shown as–Log10(permutation-based P-value). Red line shows P = 0.05. (d) Variance explained by QTLs in a cross-validation framework, compared to a null model with random genomic segments of similar size. Error bars represent the standard errors (SE) of the median of the variance explained in the 100 training/test sets.

Fig 3. Height variance explained by chromosomes.

(a) Height variance explained by each chromosome ± SE is plotted against chromosome length. Red line shows the linear fit. (b) Height variance explained by each chromosome ± SE is plotted against the maximum LOD score on that chromosome. (c) Histogram of variance explained by each of 2200 chromosomes simulated under an infinitesimal model. Red arrow shows variance explained by chromosome 14 in the real data.