Gene position scoring within transcription regulation networks - - PowerPoint PPT Presentation

Gene position scoring within transcription regulation networks

Ivan Junier, Joan Hérisson, Mohamed Elati, François Képès

Programme d’Épigénomique, Évry, France

Outline

Why positions? How to score? Which outcome?

• E. coli : conserved paired genes and their relative position

Evolutionarily conserved gene pairs Wright et al., PNAS, 104, 10559 –10564, 2007

Evolutionarily conserved gene pairs Wright et al., PNAS, 104, 10559 –10564, 2007

• E. coli : conserved paired genes and their relative position

Ori Ter

Gene position <--> Transcription regulation

Co-regulation and gene position Periodic positioning of genes regulated by the same TFs 1 D co-regulation is frequent in prokaryotes

Rapid search hypothesis TF binding site TF gene regulated TU 3D co-localization --> rapid search hypothesis

Képès, JMB, 2004

Grid interval : 92.8 kbp

Co-regulation and gene position Periodic positioning of genes regulated by the same TFs in yeast

Grid interval : 15.5 kbp Grid interval : 15.5 kbp Grid interval : 7.75 kbp

Képès, JMB, 2003

Why positions are important? Spatial co-localization

Képès, Vaillant, ComplexUS, 2003

Conceptual framework

Jackson et al., Mol. Biol. Cell, 1998

2 µm

Cook et. al, Nature, 2002

Eukaryotic nucleus

Cabrera, Jin, JMB, 2004 1µm

Bacterial cells

Experimental facts

Why positions are important? Polymer theory

• I. Junier, O. Martin, F. Képès, submitted to
• Biophys. J.

Why positions are important? Polymer theory

Periodic Random

• I. Junier, O. Martin, F. Képès, submitted to
• Biophys. J.

How to detect periodicity?

1) What is periodic? 2) Noise

Blank sites / False negatives Genes out of the periodicity / False positives + fluctuations around the original sites

Solenoidal framework

Periodicity detection = clustering detection

Principle : better score than

Play with the period : spectrum (score vs. period)

0.5 1 1 2

Statistics of circular distributions

i

j X

Binomial:

ρN,|j−i|(X = x) = C|j−i|

N−1x|j−i|−1(1 − x)N−|j−i|−2

0.5 1 1 2

ρ11,3(x)

x

Final score

S({x}) = 1 N

• i

Si({xij})

Pair score:

s(xij) = − log[pv(xij|ρN,|j−i|)]

Gene score (sum up the n first neighbors):

Si({xij}) = 1 n

(i+n)%N

• j=(i+1)%N

s(xij)

Exemple

Clustering spectrum Discrete Fourier spectrum

• I. Junier, J. Hérisson, F. Képès, to be submitted

40 55 100

100 200

2 4 6

50 40

Score Period

50 100 150 200 250 25 50 75

Amplitude Period

Interdistance

50 100 150 200 250 0.2 0.4 0.6

Period Amplitude

Positions

Some results in E. coli

1×10

5

2×10

5

Period

• 2
• 1

Score CRP

• 9510

p-value : 10-3 19020 p-value : 10-4 28000 p-value : 10-3

Unveiling chromosomal structures

• I. Junier, J. Hérisson, F. Képès, in preparation

CRP binding sites (RegulonDB 2008) : 160 targets ( 450 genes) --> 90 strong evidence

∼

Some results in E. coli

Unveiling functional relation between TFs : inference of transcription regulation

• J. Hérisson, I. Junier, F. Képès, in preparation

1×10

5

2×10

5

Period

• 2
• 1

Score CRP

1×10

5

2×10

5

Period

• 2
• 1

Score CRP CRP-OxyR

1×10

5

2×10

5

Period

• 2
• 1

Score CRP CRP-PhoB

• CRP binding sites (RegulonDB 2008) : 160 targets ( 450 genes) --> 90 strong evidence

∼

Positional score of a site (binding sites, genes,...)

Needs to specific with respect to which TF

S S∗

1×10

5

2×10

5

Period

• 2
• 1

Score CRP

• Spos = f(|S∗ − S

S |) × g(max(S∗, S))

Combining scores : learning machine technique

with J. hérisson, M. Elati, F. Képès

Biological Data

Spos Sseq Sglobal

Conclusion

How to score?

Method based on a solenoidal framework + clustering detection => valuable information for finding repetitive patterns

Which outcome?

Structural information about spatial organization of chromosomes Predicting functional relation between genes

Why positions?

Biological data show regular pattern --> space co-localization transcription regulation