Fig. 6: The pairwise modularity of NHRs.

a Heatmap depicting perturbation-perturbation similarity of DEG profiles for the NHR perturbations. The perturbation-perturbation similarity was defined by cosine similarity of the filtered log2(FC) profile. The filtered log2(FC) was derived by masking the log2(FC) values of genes that are not called as DEGs (FDR < 0.1, FC > 1.5) to zero. b Visualization of gene expression changes in selected NHR pairs. The gene expression change was measured by the corrected Wald statistic. Rows are the union DEGs of these selected NHR perturbations. c Randomization test of the pairwise modularity of NHR gene family. The schematic shows the design of the randomization test. Histogram shows the average silhouette score of pairs in 10,000 randomizations. The red line indicates the observed score from real data. The NHR GRN was randomized by swapping the network edges while preserving the network structure and properties, such as in- and out-degrees (Supplementary Methods). The gene-gene correlation was not preserved in this randomization to fully randomize the GRN. d Heatmap depicting protein sequence similarity (percent identity) for the DNA binding domain (DBD) of NHRs. The heatmap was clustered using distance matrix generated by Clustal Omega online tool from EMBL-EBI⁸⁷ (Supplementary Methods). Scatter plots showing the comparison between perturbation-perturbation similarity and sequence similarity of DBD (e) and full-length protein (f). Each data point indicates a pair of NHRs and selected pairs are labeled. g Scatter plot and randomization test for the associations between perturbation-perturbation similarity and NHR coexpression. Coexpression was measured based on the median Pearson Correlation Coefficient (PCC) of nhr gene expression in a compendium of C. elegans gene expression data across various conditions⁶⁸ (Supplementary Methods). The median coexpression level of pairs with a cosine similarity greater than 0.2 (red region in (g)) is calculated and compared with that from randomized data (Supplementary Methods). The histogram shows that the median from real data (red line) is significantly greater than that from randomized data, indicating a statistically significant association between NHR coexpression and perturbation-perturbation similarity. Source data are provided as a Source Data file.