Fig. 6: CNE prediction depends on conservation state.

a Heatmap for the ConsHMM conservation state model from ref. ³². Rows correspond to conservation states that were previously clustered into eight groups labeled at bottom and colored accordingly (Fig. 1b). For each state, the left and right halves indicate the probability of the species of the column having a nucleotide aligning to and matching to the human reference genome, respectively. Major groups of species are colored and labeled. b The first column reports the genome percent of each state. The second column contains the AUC of the CNEP score for predicting CNE bases in each state (see “Methods” section), where for this and the remaining columns the constrained elements are from PhastCons. The next column reports the AUC when exons are first extended by 200 bp. The next three columns contain the fold enrichment for CNE bases, Low_CNE bases, and the ratio of the Low_CNE to CNE enrichments. The following three columns contain the fold enrichment for notCNE bases, High_notCNE bases, and the ratio of the High_notCNE to notCNE enrichments. The last column shows the average CSS-CNEP score in CNE bases in the state. Adjacent pairs of columns with a red-white color scale are on the same color scale, while other columns are on a column specific color scale. The bottom row gives the genome coverage percent of the column. Results based on all the constrained element sets is in Supplementary Fig. 22. c ROC curves for the CNEP score identifying PhastCons CNE bases in specific ConsHMM conservation states colored based on the coloring in a. d Plot showing the AUC values for each ROC curve shown in c with the same coloring. The AUC values are displayed from left to right based on decreasing values and positioned along the x-axis based on the cumulative fraction of PhastCons CNE bases that they cover. States with the highest AUC values are labeled. Similar plots, but for additional constrained element and based on excluding bases within 200 bp of exons from the positives are in Supplementary Figs. 23 and 24. Source data are provided as a Source Data file.