Extended Data Fig. 5: Performance of LDGM precision matrices in cross-validation.

For each LD block on chromosome 22 (n = 20 LD blocks), we randomly split the 1000 Genomes EUR haploid samples into two subsets of equal size. We computed an LDGM precision matrix from one of the two subsets (the LDGM was constructed from all samples in 1000 Genomes). We computed the MSE for three comparisons: the precision matrix vs. the correlation matrix from the same sample; the precision matrix vs. the correlation matrix from the opposite sample; and the correlation matrix from one sample vs. the correlation matrix from the opposite sample. The lower whisker, lower hinge, center, upper hinge and upper whisker correspond to (lower hinge − 1.5× interquartile range (IQR)) and the 25th percentile, median, 75th percentile, and (upper hinge + 1.5× IQR), respectively.