Genomic analysis of the hierarchical structure of regulatory networks

Yu and Gurstein 1010.1073/pnas.0508637103

Supporting Information

Files in this Data Supplement:

Supporting Text
Supporting Figure 7
Supporting Figure 8
Supporting Figure 9
Supporting Figure 10
Supporting Figure 11
Supporting Figure 12
Supporting Figure 13
Supporting Figure 14
Supporting Table 2
Supporting Figure 15
Supporting Figure 16
Supporting Figure 17
Supporting Table 3

Supporting Figure 7

Fig. 7. The regulatory hierarchy built by both longest path approaches show similar results. (I) The regulatory hierarchy built by longest path of shortest distances shows similar results. (A) Summary of the hierarchy. The distribution of the number of TFs at each level has a pyramidal shape, but the distribution of the average number of targets is not monotonic. There are 13 levels in the hierarchy with much fewer TFs at each level. To have better statistical significance, we binned different levels into four groups, corresponding to the four levels that we have in the main text: 1, [1:2); 2, [2: 8]; 3, [9:10); and 4, [10:12]. (B) Fraction of essential genes. Top TFs are less essential. (C) Average number of affected genes in knockout experiments. (D) Fraction of genes that are homologs to human cancer genes. (II) The regulatory hierarchy built by longest path of maximum distances shows similar results. (A) Summary of the hierarchy. The distribution of the number of TFs at each level has a pyramidal shape, but the distribution of the average number of targets is not monotonic. There are 40 levels in the hierarchy with much fewer TFs at each level. To have better statistical significance, we binned different levels into four groups, corresponding to the four levels that we have in the main text: 1, [1:2); 2, [2: 20]; 3, [21:37); and 4, (37:40]. (B) Fraction of essential genes. Top TFs are less essential. (C) Average number of affected genes in knockout experiments. (D) Fraction of genes that are homologs to human cancer genes.

Supporting Figure 8

Fig. 8. An illustrative example to show that the top-down approach does not match our intuition in social hierarchies. (A) The bottom-up method (i.e., BFS-level) puts the clerks at the bottom level with assistants. (B) The top-down method puts the owner of the drug store at the top level with the president. The clerks are at the same level with the secretaries of the nation.

Supporting Figure 9

Fig. 9. The regulatory hierarchy built by the top-down approach shows similar results. (A) Summary of the hierarchy. There are six levels in the hierarchy with much fewer TFs at each level. To have better statistical significance, we binned different levels into the following three groups: 1, [1:2]; 2, [3: 5]; and 3, (5:6]. (B) Fraction of essential genes. Top TFs are less essential. (C) Average number of affected genes in knockout experiments. (D) Fraction of genes that are homologs to human cancer genes.

Supporting Figure 10

Fig. 10. The regulatory hierarchy built by BFS without loops shows similar results. (A) Summary of the hierarchy. The distribution of the number of TFs at each level has a pyramidal shape, but the distribution of the average number of targets is not monotonic. (B) Fraction of essential genes. Top TFs are less essential. (C) Average number of affected genes in knockout experiments. (D) Fraction of genes that are homologs to human cancer genes. In C and D, we combined levels 2, 3, and 4 as bin 2.

Supporting Figure 11

Fig. 11. The regulatory hierarchy built by TF percentage shows similar results. (A) Summary of the hierarchy. The distribution of the number of TFs at each level has a pyramidal shape, but the distribution of the average number of targets is not monotonic. To have better statistical significance, we binned TFs with different percentages into 4 groups, corresponding to the following four levels that we have in the main text: 1, 0; 2, (0:0.001); 3, [0.001:0.05); and 4, [0.05:1]. (B) Fraction of essential genes. Top TFs are less essential. (C) Average number of affected genes in knockout experiments. (D) Fraction of genes that are homologs to human cancer genes. In C and D, we combined levels 2, 3, and 4 as bin 2.

Supporting Figure 12

Fig. 12. The regulatory hierarchy shows similar results even when 20% of edges are randomly deleted. (A) Summary of the hierarchy. The distribution of the number of TFs at each level has a pyramidal shape, but the distribution of the average number of targets is not monotonic. (B) Fraction of essential genes. Top TFs are less essential. (C) Average number of affected genes in knockout experiments. (D) Fraction of genes that are homologs to human cancer genes.

Supporting Figure 13

Fig. 13. The regulatory hierarchy shows similar results even when additional 20% of edges are randomly inserted. (A) Summary of the hierarchy. The distribution of the number of TFs at each level has a pyramidal shape, but the distribution of the average number of targets is not monotonic. (B) Fraction of essential genes. Top TFs are less essential. (C) Average number of affected genes in knockout experiments. (D) Fraction of genes that are homologs to human cancer genes.

Supporting Figure 14

Fig. 14. The regulatory hierarchy shows similar results even when 20% of edges are randomly rewired. (A) Summary of the hierarchy. The distribution of the number of TFs at each level has a pyramidal shape, but the distribution of the average number of targets is not monotonic. (B) Fraction of essential genes. Top TFs are less essential. (C) Average number of affected genes in knockout experiments. (D) Fraction of genes that are homologs to human cancer genes.

Supporting Figure 15

Fig. 15. Random networks generated based on the E. coli regulatory network show similar properties to those from the yeast regulatory network.

Supporting Figure 16

Fig. 16. Average betweenness at each level of the E. coli hierarchy. P values are calculated using the Student t tests to compare the average betweenness of the top and bottom TFs with that of the middle-level TFs, respectively.

Supporting Figure 17

Fig. 17. The organizational charts for the social network examples used in the work. Note that each of these four examples represents one type of motif, which is highlighted in red. The URLs of the original organizational charts are also provided in the figure.

Table 2. Hierarchy of E. coli regulatory network

Level	Genes
4	yhiW gntR soxR cspE
3	oxyR lrhA cspA yhiX rob marR soxS exuR
2	yhiE fur crp lrp metJ cytR tdcR flhC rhaR gutM narL himA rpoS feaB cysB fis B2087 cpxR flhD rcsB rpoN fruR fhlA glnG marA nac fnr srlR dnaA rpoE uxuR modE himD hns ompR galR arcA mlc feaR lysR rhaS phoP pdhR fadR
1	metR appY trpR tyrR argR glcC xylR purR rpiR gals lldR mtlR malT atoC malI hydG emrR hycA cadC asnC yeiL idnR ilvY hupB betI uidR lexA rpoH gcvA fucR hcaR B2531 ada melR yiaJ glpR rcsA fliA cynR putA cbl dsdC treR arsR nagC csgD tdcA rtcR farR phoB araC hupA hipB yhhG fecI iclR B2090 torR caiF sdiA uhpA yjdG xapR evgA nadR adiY narP B1399 deoR gcvR acrR leuO ygaE envY alpA pspF ylcA hyfR yjbK ebgR kdpE yhdM slyA ygaA lacI rbsR nhaR mhpR birA

Table 3. Functional annotations of top-level TF in the yeast regulatory network from SGD

Gene name	ORF name	Functional annotation from SGD
SPT23	YKL020C	DNA binding and transcriptional activator activities; ER membrane protein involved, with its homolog Mga2p, in regulation of OLE1 transcription; inactive ER form dimerizes, and one subunit is then activated by ubiquitin/proteasome-dependent processing followed by nuclear targeting
HIR3	YJR140C	Transcriptional corepressor involved in the cell cycle-regulated transcription of histone genes HTA1, HTB1, HHT1, and HHT2; involved in position-dependent gene silencing and nucleosome reassembly
ADA2	YDR448W	Transcription coactivator, component of the ADA and SAGA transcriptional adaptor/HAT (histone acetyltransferase) complexes
GAT1	YFL021W	Transcriptional activator of genes involved in nitrogen catabolite repression, member of the GATA family of DNA-binding proteins; activity and localization regulated by nitrogen limitation and Ure2p
NGG1	YDR176W	Transcriptional regulator involved in glucose repression of Gal4p-regulated genes; component of transcriptional adaptor and histone acetyltransferase complexes, the ADA complex, the SAGA complex, and the SLIK complex
DAT1	YML113W	DNA binding protein that recognizes oligo(dA)•oligo(dT) tracts; Arg side chain in its N-terminal pentad Gly-Arg-Lys-Pro-Gly repeat is required for DNA-binding; not essential for viability;negative regulation of transcription from RNA polymerase II promoter
MOT3	YMR070W	Nuclear transcription factor with two Cys2-His2 zinc fingers; involved in repression of a subset of hypoxic genes by Rox1p, repression of several DAN/TIR genes during aerobic growth, and repression of ergosterol biosynthetic genes
GZF3	YJL110C	GATA zinc finger protein and Dal80p homolog that negatively regulates nitrogen catabolic gene expression by competing with Gat1p for GATA site binding; function requires a repressive carbon source; dimerizes with Dal80p and binds to Tor1p