Fig. 4: Association and prediction of CO distribution with genomic and epigenomic features.

A The non-linear correlation matrices show the comparison of pairwise features along chromosome arms, with differences in colour and size according to the correlation scale. Col_COs, Ler_COs and Hybrid_COs (CO landscapes in Col, Ler, and F2 hybrids), SNPs (SNPs density between Col and Ler), Seq_div (sequence diversity in the population of 2029 Arabidopsis accessions), INV_TRANS (inversions and translocations between Col and Ler), BrdU_labelled (origins of DNA replication, log₂(BrdU/gDNA)), SPO11 (SPO11-1-oligos, log₂(oligos/gDNA)), Genes, TEs and GC (gene, TE and GC density), ATAC and DNase (chromatin accessibility, ATAC-seq and DNase-seq, log₂(Tn5/gDNA) and log₂(DNase/gDNA)), H3K4me1/2/3, H3K9me2, H3K27me1 (euchromatin, heterochromatin, and Polycomb histone marks, ChIP-seq, log₂(ChIP/input)), REC8 (cohesin, ChIP-seq, log₂(ChIP/input)), mCG, mCHG and mCHH (DNA methylation in CG, CHG, and CHH contexts, proportion methylated cytosine), MNase (nucleosome occupancy, MNase-seq, log₂(MNase/gDNA)). The importance of each of the 14 features for explaining variation in CO distribution at the chromosome-arm (B) and genome scale (D), respectively. The size of points corresponds to the importance. The cumulated proportion of variation that can be explained with the features at the chromosome-arm (C) and genome scale (E), respectively. The top six and five most important features, for which the cumulative proportion of variation that can be explained reaches the plateau, are coloured separately. F The chromosomal distribution of observed and predicted COs. The CO profiles of individual chromosomes were predicted using profiles of the top five most important features from the other four chromosomes. The Spearman’s and non-linear correlation coefficients between the predicted and observed CO distributions for each chromosome and the whole genome are indicated, respectively. Source Data are provided as a Source Data file.