Fig. 3: Structural variants occur at different frequencies in populations and are non-randomly distributed across the genome.

a, Frequencies of inversions (inv), duplications (dup) and translocations (trans) by population. European hemp, Asian hemp and MJ populations differ significantly in average translocation and duplication counts, but not in inversions. Each box plot represents the median (centre line), two hinges (quartiles) and two whiskers (1.5× the interquartile range (IQR). b, Non-random genomic distribution of translocations (purple histograms), duplications (dark red bands) and inversions (mapped as length-scaled yellow bars on the right side of chromosomes, with each bar equal to one inversion). c, LD plot limited to 200-kb interactions, highlighting the general decay curves, with the X chromosome exhibiting a markedly reduced decay rate. Collectively, across the 78 haplotype-resolved, chromosome-scale assemblies, LD decay plots showed decay to half of the maximum r² value around 10 kb, which is similar to wild outcrossing soy and rice populations. d, LD decay plots extended to 800 kb. The increase in long-range (hundreds of kb to Mb) LD patterns found in certain SNP pairs further underscored the importance of using accurately phased genome assemblies and consideration of SVs for mapping and improvement efforts.