HAL Id: hal-02797859
https://hal.inrae.fr/hal-02797859
Submitted on 5 Jun 2020
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Bursting in gene expression model
Romain Yvinec, Michael C. Mackey, Marta Tyran-Kamińska, Changjing Zhuge
To cite this version:
Romain Yvinec, Michael C. Mackey, Marta Tyran-Kamińska, Changjing Zhuge. Bursting in gene
expression model. Colloque PDMP, Agence Nationale de la Recherche (ANR). FRA., May 2015,
Saint Martin de Londres, France. �hal-02797859�
Bursting in gene expression model
Romain Yvinec 1 , Michael C. Mackey 2 , Marta Tyran-Kami´ nska 3 , Changjiing Zhuge 4 , Jinzhi Lei 4
1
BIOS group, INRA Tours, France.
2
McGill University, Canada.
3
University of Silesia, Poland.
4
Tsinghua University, China.
Goodwin’s deterministic model
Stochastic gene expression model
Analytical results on a reduced model
Ongoing work 1) Taking into account division
Ongoing work 2) Recovering the burst statistics
Crick (1958) : Central Dogma.
Jacob, Perrin, S´ anchez, Monod (1960) : Operon.
$
’ ’
’ ’
’ ’
&
’ ’
’ ’
’ ’
% dM
dt “ λ
1pE q ´ γ
1M , dI
dt “ λ
2M ´ γ
2I , dE
dt “ λ
3I ´ γ
3E .
(1)
Goodwin (1965), Griffith (1968), Othmer (1976), Selgrade (1979)...
Goodwin’s model Stochastic model
The transcription rate function λ 1
§ Inducible Operon : Repressors R interacts with both the Operator O and the Effector E ,
R ` nE é K
1RE n , K 1 “ RE n
R ¨ E n , O ` R é K
2OR , K 2 “ OR
O ¨ R . With QSSA, and if O tot ! R tot ,
λ 1 pE q „ O O tot
“ 1 ` K 1 E n
1 ` K 2 R tot ` K 1 E n . (2)
§ Repressible Operon : Similar but
O ` RE n K é
2ORE n , K 2 “ ORE n
O ¨ RE n . and we get
λ 1 pE q „ O O tot
“ 1 ` K 1 E n
1 ` pK 1 ` K 2 R tot qE n . (3)
Bifurcation analysis in ODE
0 5 10 15 20
0 2 4 6 8 10
K κd
uninduced bistable induced
§ Inducible : Mono-stability or Bi-stability.
§ Repressible : Mono-stability or limit cycle.
$
’ ’
’ ’
&
’ ’
’ ’
% dx
1dt “ γ
1rλ
1px
3q ´ x
1s , dx
2dt “ γ
2px
1´ x
2q , dx
3dt “ γ
3px
2´ x
3q . (4) Here λ
1pxq “ κ
d1 ` x
nK ` x
n.
0 1 2 3 4 5 6 7
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
x
Goodwin’s model Stochastic model Reduced model With division Inverse problem
Eldar and Elowitz (Nature 2010)
’New’ Central dogma
DNA
mRNA Protein
ON
OFF
Novick & Weiner (1957),Ko et al. (1990),Ozbudak et al. (2002),Elowitz et al. (2002),Raser & O’Shea (2004)...
$
’ ’
’ ’
’ &
’ ’
’ ’
’ %
dx
dt “ G ptqλ 1 pyptqq ´ γ 1 xptq , dy
dt “ λ 2 xptq ´ γ 2 yptq , pG “ 0q ÝÝÝÝá âÝÝÝÝ αpyptqq
βpyptqq
pG “ 1q .
(5)
Rigney & Schieve (1977),Berg (1978),Peccoud & Ycart (1995),Thattai & Van Oudenaarden (2001)...
Outline
Goodwin’s deterministic model
Stochastic gene expression model
Analytical results on a reduced model
Ongoing work 1) Taking into account division
Ongoing work 2) Recovering the burst statistics
Can we perform a systematic bifurcation theory on such systems ?
§ We are interested in long time behavior.
§ We want to know how many modes has the stationary distribution.
§ This requires in practice ’analytical’ solution.
Goodwin’s model Stochastic model Reduced model With division Inverse problem
A subclass of the ’three-stage’ (DNA, mRNA, Protein) model is the 1D-bursting model (Storage model)
Lf pxq “ ´γpxqf 1 pxq ` λpxq ż 8
0
pf px ` yq ´ f pxqqhpx, yqdy (6) where h is the burst size distribution, ş 8
0 hpx, yqdy “ 1. If hpx, yq “ ´ ν
1νpxq px`y q , ν Œ and ν Ñ 8 0,
Lf pxq “ ´γ pxqf 1 px q ` λpxq νpxq
ż 8
x
f 1 pz qν pzqdz (7) Any stationary distribution satisfies ş 8
0 Lf pxqu ˚ pxqdx “ 0, so that ż 8
0
”
´ γ pxqu ˚ pxq ` νpxq ż x
0
λpy q
νpyq u ˚ py qdy ı
f 1 pxqdx “ 0 . (8) Hence
u ˚ pxq “ νpxq C γpxq exp
´ ż x λpy q γ py q dy
¯
(9)
Bifurcation analysis in SDE
0 1 2 3 4 5 6 7
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
x
§ Inducible : Uni-modal or Bi-modal.
§ Repressible : Uni-modal.
(under technical assumptions...) upt, xqdx “ P xptq P rx, x ` dx q ( converges as t Ñ 8 (in L 1 ) towards u ˚ ,
du
˚dx “
„
λpxq ´ γp1 ` x bpxq q
u
˚pxq γx . with bpxq “ ´
ννpxq1pxq. and γpxq “ γx.
0 2 4 6 8 10 12 14 15
0 5 10 15 20
K
The presence of bursting can drastically alter regions of bistability
κd−/+
b*κb=κd =K bκb−/+ ; b=0.01 bκb−/+ ; b=0.1 bκb−/+ ; b=1 bκb−/+ ; b=10
n=4
κd ≡ b*κb
κd−/+
Goodwin’s model Stochastic model Reduced model With division Inverse problem
Stationary distribution p λ, γ, b q ñ p u ˚ q
3 3.5 3.8 4 4.44 5 5.5
0 1 2 3 4 5 6 7 8
6
x κb
2.8 3 3.15 3.3 3.7 4 4.45
0 1 2 3 4 5 6 7 8
5
x κb
Here λpxq “ κ b 1 ` x n
K ` x n and bpxq “ b (νpxq “ expp´x{bq).
From the 3-stage to the bursting model
$
’ ’
’ ’
’ &
’ ’
’ ’
’ %
dx
dt “ G ptqλ 1 pyptqq ´ γ 1 xptq , dy
dt “ λ 2 xptq ´ γ 2 yptq , pG “ 0q ÝÝÝÝá âÝÝÝÝ αpyptqq
βpyptqq
pG “ 1q .
(10)
§ If the mRNA lifetime si short (γ 1 Ñ 8), we can perform an adiabatic reduction (xptq « Gptq λ γ
11
pyptqq) :
$
’ &
’ %
dy
dt “ Gpt q
λ2γλ11
py ptqq ´ γ
2y pt q , pG “ 0q ÝÝÝÝá âÝÝÝÝ
αpyptqqβpyptqq
pG “ 1q .
(11)
Goodwin’s model Stochastic model Reduced model With division Inverse problem
$
’ &
’ %
dy
dt “ G ptq
λγ2λ11
py pt qq ´ γ
2ypt q , pG “ 0q ÝÝÝÝá âÝÝÝÝ
αpyptqqβpyptqq
pG “ 1q .
(12)
§ If the Gene active periods are short (β Ñ 8), we obtain the bursting model
dy
dt “ Z ptq ´ γ
2y ptq , (13) where Z “ ř
i Z i δ T
iis a jump process, of jump rate αpyptqq and jump size cumulative distribution of separated form
P ypT
i`q ě z | y pT
i´q “ y (
“ exp
´
´ ż
zy
γ
1β λ
1λ
2pw qdw
¯ .
Remark
For a constitutive gene, b :“ λ γ
1λ
21
β is the average number of
proteins produced per Gene activation event.
Remark
For the ’2-stage’ model (Telegraph),
$
’ &
’ %
dx
dt “ G ptqλpxptqq ´ γ xptq , pG “ 0q ÝÝÝÝá âÝÝÝÝ αpxptqq
βpxptqq
pG “ 1q . (14)
we have (See Boxma et al. 2005) du ˚
dx “
” αpxq
γx ´ βpxq λpxq ´ γ x
´ λpxq{x ´ γ ` γxpλ 1 pxq ´ γq{pλpxq ´ γ xq λpxq
ı
u ˚ (15)
Small recap’ on the forward problem (1/2)
§ We performed an adiabatic reduction to make the problem analytical tractable.
§ We solved the reduced problem for arbitrary coefficients at equilibrium.
§ We performed an (deterministic-analogous) bifurcation study.
0 2.5 5
0 0.5 1 1.5
X
density
0 2 4 6 8 10 0
0.2 0.4 0.6
Y
density
0 5 10
0 0.2 0.4 0.6 0.8 1
X
0 2 4 6 8 10 0
0,1 0,2
Y 0 2 4 6 8 10
0 0,1 0,2
Y
0 25 50
0 0.2 0.4 0.6 0.8 1
X
0 2 4 6 8 10 0
0,1 0,2
Y
0 200 400
0 0.2 0.4 0.6 0.8 1
X
γ1=0.1 γ1=1 γ1=10 γ1=100
Small recap’ on the forward problem (2/2)
§ Careful ! The two notions of deterministic bistability and
’stochastic bistability (bimodality) are in fact quiet different
0 100 200 300 400 500 600 700 800 900 1000
0 0.5 1 1.5 2
2.5 Deterministic Bistabilty
Time x
κd=2.3 γ=1 K=4 n=4
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 0
2 4 6 8 10 12
14 Bistability with Bursting Only
Time x
§ (mean) Switching time : can quantify the ’stability’ of each
state.
Outline
Goodwin’s deterministic model
Stochastic gene expression model
Analytical results on a reduced model
Ongoing work 1) Taking into account division
Ongoing work 2) Recovering the burst statistics
Similar results may be obtained for a ’bursting-division’ model.
Lf pxq “ d pxq ż x
0
pf pyq ´ f pxqqκpx, yqdy
` λpxq ż 8
0
pf px ` yq ´ f pxqqhpx, y qdy For instance, with uniform repartition kernel (κpx, yq “ 1{x), constant division rate d and constant exponential burst size (hpx, y q “ expp´y{bq),
d dy u ˚ “
„
´ λ 1 py q ` d
λpy q ` d ` λpy q λpy q ` d
´ 1 x ` 1
b
¯
´ xb 2 bx ` 1 ´ 1
x
u ˚ pyq
Goodwin’s model Stochastic model Reduced model With division Inverse problem
This may be used to predict the long time behavior of a dividing cell population
scenario 1
A cell tree
Protein
-1-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 0
0.5
Histogram and analytical solution 1
scenario 2
A cell tree
Protein
-1-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 0
0.5
Histogram and analytical solution 1
Outline
Goodwin’s deterministic model
Stochastic gene expression model
Analytical results on a reduced model
Ongoing work 1) Taking into account division
Ongoing work 2) Recovering the burst statistics
Goodwin’s model Stochastic model Reduced model With division Inverse problem
Inverse Problem : p u ˚ q ñ p λ, γ, b q
For a constitutive gene, we can infer the burst rate (in protein lifetime unit) λ γ and the mean burst size b from the first two (stationary) moments
bλ γ “ E
“ X ‰ ,
b “ Var pX q E
“ X ‰ .
For an auto-regulated gene, we can inverse the formula for the stationary pdf :
pxu ˚ pxqq 1
u ˚ pxq “ λpxq
γ ´ x
bpxq .
Simulated data
Density reconstruction by Kernel Density Estimation
x
0 2 4 6 8 10 12 14 16 18 20
0 0.05
0.1 0.15 0.2 0.25
True solution Kernel density estimation Histogram
Goodwin’s model Stochastic model Reduced model With division Inverse problem
Inferred bursting rate
protein
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
bursting rate
0 2 4 6 8
10 Recovering Burst rate
Estimated rate with the true b True solution
Estimated rate with a wrong b
Resulting Probability Density Function
protein
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Density
0 0.1 0.2 0.3
Resulting density
True solution KDE Fit with the true b Fit with the wrong b
Goodwin’s model Stochastic model Reduced model With division Inverse problem
Single cell data on self-regulating gene
§ Synthetic Tet-Off in budding yeast.
§ Feedback modulated by an external parameter
(doxycycline)
1) Kernel Density Estimation
YFP Reporter
1 1.5 2 2.5 3 3.5 4 4.5 5
Goodwin’s model Stochastic model Reduced model With division Inverse problem
2) Finding the ’best’ mean burst size (KL distance)
log
10(b)
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 lo g
10(e rr o r) K L
-6 -4 -2 0 2
KL-Error as a fonction of b
3) Inferred burst rate
protein
1 1.5 2 2.5 3 3.5 4 4.5 5
Burst rate
0 50
100 Estimating burst rates
Goodwin’s model Stochastic model Reduced model With division Inverse problem