Ohnologs and Regulatory Networks

1
Ohnologs and Regulatory Networks

• Robbie Sedgewick
• Group Meeting
• March 2, 2006

2
Ohnologs paralogs that arose through
polyploidization

• Ohnologs taken from YGOB (Byrne Wolfe 2005)
• 554 Ohnolog pairs
• 1108/6540 17 of yeast genome is an ohnolog
• Relatively complete dataset.
• Sparse graph

3
Regulatory Network

• Harbison et al 2004.
• Used CHIP to identify binding locations for 203
  TFs
• Of those, 102 TFs had enough hits to determine
  TF binding sites (motifs) computationally with
  constraints on binding strength and (optionally)
  conservation.
• Noisy.
• Not complete.

4
Basic facts about the regulatory data.
p lt 0.001 P lt 0.005
No conservation Loose 9778 regulations
Conserved in 2 yeasts
Conserved in 3 yeasts Strict 3328 regulations
5
Duplicated transcription factors
Ohnologs obs (exp) Paralogs obs (exp)
102 TFs Count 37 (15) 51 (40)
102 TFs Pairs 14 27
203 TFs Count 68 (30) 110 (80)
203 TFs Pairs 26 81

• Number of Ohnologs expected to be TFs
  972102/6540 15

In some cases, only one member of a pair is
considered a TF in Harbison et al data.
6

• For more ohnolog and paralog stats with
  correlation coefficients
• http//goby.compbio.cs.cmu.edu/DurandWiki/index.ph
  p/Ohnolog_and_paralog_pairs_that_are_transcription
  _factors

7
Combining ohnology and regulation

• g1 is significantly similar to g3

g1 regulates g2
8
Null hypotheses
Compare with paralogs
Compare with randomized graphs
9
Paralogs

• Use sequence similarity to determine Paralogy.
• Eval cutoff of 10-10. (soon to use NC)
• 8572 paralogous pairs
• Dense (compared to ohnologs)
• Error prone

10
Randomization method

• Method due to George and Robbie
• Take two networks and scramble the name mapping
  between them.
• Results in random combination of graphs that
  preserves network properties (e.g., node degree)
  of both component graphs.

11
Randomization method
(name lookup)
(scrambled lookup table)
12
Results

• How often is a gene regulated by both members of
  an ohnolog pair?
• How often are both members of an ohnolog pair
  regulated by the same TF?
• Geometric motifs

13
How many genes are regulated by
both members of an ohnolog pair?
only one member of an ohnolog pair?
Note a target may be counted more than once if
regulated by more than one ohnolog pair.
For comparison, how many ohnolog pairs regulate
at least one common target?
14
Strict
Num Randav p-val
Ohnologs 989 1180 10-4
Paralogs 2502 3080 9.5x10-3
Triangles Num Randav p-val
Ohnologs 109 12. 10-4
Paralogs 257 25 10-4
Ohnolog pairs that regulate at least one common
target
Pairs Num Randav p-val
Ohnologs 10 2.8 10-4
Paralogs 16 5.3 10-4
15
Loose
Ohnolog pairs that regulate at least one common
target
16
Results

• How often is a gene regulated by both members of
  an ohnolog pair?
• How often are both members of an ohnolog pair
  regulated by the same TF?
• Geometric motifs

17
How often does a transcription factor regulate
both members of an ohnolog pair?
only one member of an ohnolog pair?
For comparison, how many ohnolog pairs have at
least one regulator in common?
18
Strict
TFs that regulate both members of an ohnolog pair
Genes that regulate only one member of an ohnolog
pair
Triangles Num Randav p-val
Ohnologs 38 8.2 10-4
Paralogs 267 90 10-4
Num Randav p-val
Ohnologs 716 776 10-4
Paralogs 9994 10900 0.11
Ohnolog pairs have at least one regulator in
common
Pairs Num Randav p-val
Ohnologs 26 6.8 10-4
Paralogs 209 80 10-4
19
Loose
20
Results

• How often is a gene regulated by both members of
  an ohnolog pair?
• How often are both members of an ohnolog pair
  regulated by the same TF?
• Geometric motifs

21
Motifs and evolution
We can understand complex motifs by considering
what happens after a whole genome duplication.
22
Results one pair
Regulate each other
Regulates Partner
Num Randav p-val
Ohnologs 1 0.04 0.04
Paralogs 1 0.0439 0.0439
Num Randav p-val
Ohnologs 4 0.52 6x10-4
Paralogs 8 1.0 2x10-4
Strict
23
Results one pair
Regulate each other
Regulates Partner
Num Randav p-val
Ohnologs 5 1.3 5.3x10-3
Paralogs 11 2.6 1x10-4
Num Randav p-val
Ohnologs 1 0.065 0.065
Paralogs 1 0.0615 0.0603
Loose
24
For pictures of these motifs

• http//www.cs.cmu.edu/bobsedge/images/all-self-re
  g.pdf
• http//www.cs.cmu.edu/bobsedge/images/all-self-re
  g-loose.pdf

25
Motifs and evolution
We can understand complex motifs by considering
what happens after a whole genome duplication.
WGD
Loss
26
Results Two pairs
1 loss
0 loss
2 loss
Strict
27
Results Two pairs
1 loss
0 loss
2 loss
Loose
28

• For a table with more complete motif stats
• http//www.cs.cmu.edu/bobsedge/images/tablemod.pd
  f

29
An example Duplicated Iron Pathway?

• AFT1 and AFT2 are TFs that are also ohnologs and
  both regulate iron deprivation response pathways.
• 26 of the 60 genes regulated by AFT2 are ohnologs
  (strict).
• Maybe iron deprivation response pathway was
  duplicated in WGD? buffering?
• AFT1 and AFT2 motif
• http//www.cs.cmu.edu/bobsedge/images/AFT2_and_fr
  iends2.pdf

30
Additional Sources of Data

• Gene Coexpression
• Synthetic Lethal interactions
• Protein-Protein interactions (Y2H)
• Domain information
• Genes that were ohnologs (singletons from YGOB)