Timing the Cell Cycle

1
Timing the Cell Cycle

• Seth Berman
• Julian Lange
• Reina Riemann
• Ezequiel Alvarez-Saavedra

2
Outline
Seth phase
Eze phase
The cell cycle
A biological model
The algorithm and results
Julian phase
Reina phase
The project
3
Cell cycle early findings

• histone mRNA oscillates during the yeast cell
  cycle (Hereford et al, 1981)
• most genes expressed at G1/S transition contain
  binding sequences for
• specific transcription activators (Koch and
  Nasmyth, 1994)
• many cell cycle-regulated genes are involved in
  processes (budding,
• cytokinesis, etc.) that occur only once per cell
  cycle

cell cycle is a complex self-regulating program
4
Background Spellman et al (1998)

• used DNA microarrays to analyze mRNA
• levels in synchronized cell cultures
• identified genes whose mRNA expression
• profiles were similar to those of genes known
• be regulated by the cell cycle

800 genes are cell cycle regulated
5
Background Simon et al (submitted)

• performed genome-wide location analysis of nine
  known cell cycle
• transcription activators
• compared data to Spellman et al microarray gene
  expression experiments

6
Background Simon et al (submitted)

• transcription activators function to regulate
  gene expression and diverse stage-specific
  functions during the cell cycle

activators also regulate expression of the other
transcription activators

• leads to temporal regulation of the cell cycle

7
Project Goal

• quantitative integration of genome-wide location
  analysis and cell cycle
• expression data to determine direct regulatory
  relationships among nine
• transcription activators
• aims
• to quantitatively validate relationships
  established by location analysis,
• with the expression data
• to optimize temporal relationships based on time
  lags in expression
• to propose a temporal model for the expression
  of the nine activators

8
Activators bind at promoters of other activators
Ace2
Swi5
Mbp1
Ndd1
Swi6
Mcm1
Swi4
Fkh1
Fkh2

• data from Simon et al (submitted), p0.001

9
Terminology

• Ace2 is a child of four parents

Ace2
Swi5
Mbp1
Ndd1
Swi6
Mcm1
Swi4
Fkh1
Fkh2
10
Cell cycle expression profiles of activators
Mcm1
Fkh2
Fkh1
Swi5
Ace2
Ndd1
Swi6
Swi4
Mbp1

• data from Spellman et al (1998)

11
Ace2 a child with four parents
12
Data processing

• naive interpolation for missing cell cycle
  expression data points
• multivariate regression models for all time lags
  for each child and parents set to investigate
  optimal time lag and combinatorial parent
  relationship
• child N(?child ?? parents, ?2)
• nested likelihood ratio tests combined with
  F-test to validate p values

13
Score

• nested likelihood ratio tests
• T(X) 2 log ((p P(childparent,H1)/(p
  P(childHo))

14
Algorithm
For each child For each time lag(0 up to
maximum time lag) For each parent score
calculate minimum score while (number of
edges in the modelltnumber of parents) If
(score lt threshold) attempt to add another
edge
15
Results initial network to be evaluated
Ace2
Swi5
Mbp1
Ndd1
Swi6
Mcm1
Swi4
Fkh1
Fkh2
16
Time lag 0 minutes
Ace2
Swi5
Mbp1
p0.003
p0.009
Ndd1
Swi6
p0.003
Mcm1
Swi4
p10-284
Fkh1
Fkh2
17
Time lag 7 minutes
Ace2
Swi5
Mbp1
p0.05
p0.00006
Ndd1
Swi6
p0.003
p0.016
Mcm1
Swi4
Fkh1
Fkh2
18
Time lag 14 minutes
Ace2
Swi5
Mbp1
p0.001
p0.001
Ndd1
Swi6
p0.01
p0.002
Mcm1
Swi4
Fkh1
Fkh2
19
Time lag 21 minutes
Ace2
Swi5
Mbp1
p0.0002
p0.00002
Ndd1
Swi6
p0.004
p0.017
Mcm1
Swi4
Fkh1
Fkh2
20
Time lag 28 minutes
Ace2
Swi5
Mbp1
p0.1
p0.02
Ndd1
Swi6
p0.001
Mcm1
Swi4
Fkh1
Fkh2
p0.000003
21
Time lag 35 minutes
Ace2
Swi5
Mbp1
p0.003
p0.05
Ndd1
Swi6
p0.08
p0.007
Mcm1
Swi4
Fkh1
Fkh2
22
Time lag 42 minutes
Ace2
Swi5
p0.005
Mbp1
p0.05
Ndd1
Swi6
p0.03
p0.007
Mcm1
Swi4
Fkh1
Fkh2
23
Time lag 56 minutes
Ace2
Swi5
Mbp1
p0.01
p0.0002
Ndd1
Swi6
p0.04
Mcm1
Swi4
Fkh1
Fkh2
p0.008
24
Significant edges
Parents
Children
Time Lag
7
Swi6 Swi4
Swi4 Ndd1 Ace2 Swi5
Swi6 Swi4
14
14
Ndd1 Fkh1 Fkh2
14
7
21
Ndd1 Fkh2 Mcm1
14
14
25
A temporal model
Swi6
Swi4
0
56
7
14
49
42
21
Swi4
Swi6
35
28
Ndd1
26
A temporal model
0
56
7
14
49
42
21
Swi5
Fkh2
Ndd1
Mcm1
35
28
Ndd1
Mcm1
Fkh2
27
A biological model
?
14-21
Swi6
Mcm1
Swi4
Ace2
G1
M
Swi5
7-14
14-21
S
G2
21
Fkh2
Fkh1
Ndd1
Mcm1
28
Conclusion

• initial integration of location and expression
  data at different time lags and proposition of a
  temporal cell cycle model
• combine information from multiple data sources
  (cdc15, cdc28, elutriation, alpha-factor arrest)
• build a more refined time model for each
  child/parent set
• iteratively update the values for the missing
  data points

Perspectives