• Aucun résultat trouvé

Bursting in gene expression

N/A
N/A
Protected

Academic year: 2021

Partager "Bursting in gene expression"

Copied!
20
0
0

Texte intégral

(1)

HAL Id: hal-02797860

https://hal.inrae.fr/hal-02797860

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

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. Colloque PDMP, Agence Nationale de la Recherche (ANR). FRA., Nov 2015, Tours,

France. �hal-02797860�

(2)

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.

(3)

Historical example of PDP : Cell cycle toy model

PDMP sur C `

R

`

, L

1

˘

Bursting gene expression

(4)

On note x

n

une quantit´ e physiologique d’une cellule aux instants de division t

n

, mod´ elis´ e par

x

n`1

“ k

´1

`

Q

´1

pQpx

n

q ` τ

n`1

q ˘

(1) o` u

Hpτ q “ P τ

n

ě τ | x

n

“ x ( , Qpz q “

ż

z

ϕpy q{g py qdy . (2)

Si x

0

a pour densit´ e f

0

P L

1

pR

`

q, alors x

n

„ f

n

o` u f

n`1

pxq “ Pf

n

pxq ,

Pf pxq “

ż

kpxq

0

κpx, y qf pyqdy , κpx, yq “ ´ d

dx H pQpkpxqq ´ Q pyqq .

(3)

On v´ erifie que si Q , k sont des bijections, absolument continues de

R

`

Ñ R

`

, et H de R

`

Ñ r0, 1s, alors P est un op´ erateur de

Markov, i.e. laisse tf P L

1

, }f } “ 1u invariant.

(5)

Theorem (Lasota, Mackey (84)) Si τ

n

a une densit´ e h, ş

xhpxqdx ă 8, et

xÑ8

lim

Q pkpxqq

Q pxq ą 1 ,

Dx

0

ě 0 t.q. hpxq ą 0, @x ą x

0

,

(4)

alors pP

n

q est asymptotiquement stable sur L

1

Exemple :

§

ϕ “ 1, g pxq “ gx

α

, kpxq “ 2x.

§

hpτ q “ Ψpτ ´ t

b

q1

τětb

pτ q

Alors Q p2xq{Qpxq “ 2

1´α

ÞÑ asympt. stable pour α ă 1.

x

n`1

“ 1 2

“ x

n1´α

` g p1 ´ αqτ

n`1

1{p1´αq

. (5)

(6)

Sur F “ E ˆ I , E Ď R

d

, | I |ă 8, ` a un PDMP Y

t

“ pX

t

, i

t

q de caract´ eristique pg , ϕ, κq, on associe le semi-groupe P ptq sur L

1

pF q solution de

Bu

Bt “ ´ Bpguq

Bx ´ ϕu ` K pϕuq , (6)

o` u K est un op´ erateur de Markov donn´ e par, @B P BpE q, u P L

1

, ż

B

Kupyqdy “ ż

E

κpy , Bqu pyqdy .

On remarque que @f P BpEq, u P L

1

, ż

E

f pyqKupyq “ ż

E

upyqK

˚

f pyqdy , o` u

K

˚

f py q “ ż

E

f pz qκpy, dzq “ E “

f pY pτ

1

qq | Y pτ

1´

q “ y ‰

.

(7)

On construit P ptq avec un th´ eor` eme de perturbation du

dt “ A

0

u ` Bu , (7)

o` u A

0

u “ ´

BpguqBx

´ ϕu g´ en` ere un semi-groupe S

0

sous-stochastique, donn´ e par

ż

B

S

0

ptqupyqdy “ ż

E

1

B

t

yqe

´

şt

0ϕpπsyqds

upy qdy ˆ

“ ż

E

1

B

t

yqH pQpπ

t

yq ´ Q pyqq upy qdy

˙

On remarque que @f P BpEq, u P L

1

, ż

E

f pyqS

0

ptqupyq “ ż

E

upyqT

0

ptqf py qdy , o` u

T

0

ptqf pyq “ f pπ

t

yqe

´

şt

0ϕpπsyqds

“ E

y

“ f pY

t

q1

tăt1

.

(8)

Th´ eor` eme de Perturbation (Philipps, Kato...)

On a (formule de variation de la constante) P pt qu “ S

0

ptqu `

ż

t

0

S

0

pt ´ sqBPpsquds (8) Si B est born´ e, ou si B : DpA

0

q Ñ L

1

, et

ż

E

Aupy q ` Bupy qdy “ 0 , @u P DpA

0

qp “ L

1ϕ

q alors on peut d´ efinir un semi-groupe ’minimal’ solution de (8), sous-stochastique, donn´ e par

P ptqu “

8

ÿ

n“0

S

n

ptqu , S

n`1

ptqu “

ż

t

0

S

0

pt ´ s qBS

n

psquds “ ż

t

0

S

0

pt ´ sqK pϕS

n

psquq ds .

(9)

Interpr´ etation : on montre que (Tyran-Kami´ nska et al. 2012) ż

B

P ptqupyqdy “ ż

E

P

y

Y ptq P B, t ă t

8

( upyqdy En effet, soit T

n

ptqf pyq “ E

y

“ f pY

t

q1

tătn`1

‰ , par la prop Markov, T

n

ptq1

B

pxq “ T

0

ptq1

B

pxq `

ż

t

0

T

0

ps q pϕK

˚

pT

n´1

ptq1

B

qq pxq . Ainsi, par dualit´ e,

ż

E

T

n

pt q1

B

px qupxqdx “ ż

B

S

0

ptqupx qdx

` ż

t

0

ż

E

T

n´1

pt ´ sq1

B

pxqK pϕS

0

psquq . Par r´ ecurrence, on a

ż

B n

ÿ

i“0

S

i

ptqupxqdx “ ż

E

T

n

ptq1

B

pxqu pxqdx .

(10)

L’approche ’thm de perturbation’ permet de donner des conditions pour obtenir un semi-groupe stochastique (autre que t

8

“ 8). Par exemple,

§

Pour tout u P L

1

, lim

1n

ř

n

k“0

pK pϕRpλ, A

0

qqq

k

u exists in L

1

§

J “ lim

λÑ0

K pϕRpλ, A

0

qqu est ergodique en moyenne.

Remarque : J est un op´ erateur stochastique associ´ e au noyau J px, Bq “

ż

8

0

κpπ

s

x, Bqϕpπ

s

xqe

´

şt

0ϕpπsyqds

i.e. `

J

˚

f pyq “ ş

E

f pzq J py , dzq ˘

(11)

Avec F “ E “ R

`

, g “ ´γpxq ă 0, et κpx, ¨q ` a densit´ e kpx, yqdy . Rpλ, Aqu “

ż

8

x

1

γpxq e

Qλpyq´Qλpxq

upy qdy o` u Q

λ

“ Q ` λG , et

Qpxq “ ż

x

ϕpz q

γpzq dz , G pxq “ ż

x

1 γpzq dz J px, dy q “ j px, yqdy , jpx, yq “

ż

y

0

1

p0,xq

pz qkpz, yq ϕpz q

γpz q e

Qpxq´Qpzq

dz Theorem (Mackey, Tyran-Kami´ nska, Y.)

Si Qp0q “ G p0q “ 8, et si l’une des deux conditions est vraie

§

J a une mesure invariant v

˚

ą 0 a.e.

§

Si k est born´ e, m

1

pxq “ ş

8

x

py ´ xqkpx, yq est loc. born´ e, et si lim sup

xÑ8

ż

x

0

ˆ

m

1

pz q ϕpz q γpz q ´ 1

˙

e

Qpxq´Qpzq

dz ă 0

alors Ppt q est un op´ erateur stochastique.

(12)

Temps long

L’´ etude en temps long repose sur des r´ esultats du type Theorem (Pich´ or, Rudnicki (2000))

Si Pptq est stochastique, partiellement ` a noyau, et poss` ede une unique densit´ e invariante alors Ppt q est asymptotiquement stable.

Partiellement ` a noyau : il existe t

0

ą 0 et p s.t.

ż

8

0

ż

8

0

ppx , y q dy dx ą 0 et Ppt

0

qupx q ě ż

8

0

ppx, y qupy q dy

Remark

Il est facile de montrer que P ptq est partiellement ` a noyau si ş ş kpx, yqϕpxqdxdy ą 0 en utilisant

Pptqu ě S

1

ptqu .

(13)

Si kpx, yq “ ´

ν1νpxqpx`yq

, ν ą 0, ν Œ et ν Ñ

8

0, alors v

˚

pxq “ ´ν

1

pxqe

´Qpxq

et u

˚

pxq “

Cγpxqνpxq

exp

´ ş

x λpyq

γpyq

dy

¯

sont des densit´ es stationnaires “candidates”.

Si Qp0q “ G p0q “ 8 et c “

ż

8

0

νpxq

γpxq e

´Qpxq

dx ă 8, et ´ ż

8

0

ν

1

pxqe

´Qpxq

dx ă 8,

alors pour toute densit´ e initiale u

0

, la solution P ptqu converge en

norme L

1

quand t Ñ 8 vers l’unique solution stationnaire donn´ ee

par u

˚

.

(14)

§

Remarque : on peut utiliser ce r´ esultat pour prouver une convergence en temps long dans des cas ’non-triviaux’.

En particulier, dans le mod` ele de croissance-fragmentation, avec Bupt, xq

Bt ` Bg pxqupt, xq

Bx “ ´ϕpxqupt , x q ` ż

8

x

ϕpyqupt , yq ν

1

pxq νpyq dy, En prenant

ϕpxq “ αx

β´1

` x

β`1

g pxq “ x

β

νpxq “ x,

pour 0 ď β ď 1, 0 ă α ă 1, on a convergence en temps long vers u

˚

pxq “ νpxq

cg pxq e

´

şx

¯ x

ϕpyq gpyqdy

,

mais ϕ

g R L

10

(15)

Application : bifurcation

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.

La densit´ e stationnaire u

˚

v´ erifie, du

˚

dx “

ϕpx q ´ 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*κbd =K b−/+ ; b=0.01 b−/+ ; b=0.1 b−/+ ; b=1 b−/+ ; b=10

n=4

κd ≡ b*κb

κd−/+

(16)

Stationary distribution p ϕ, 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).

(17)

Extension ` a un mod` ele 2d

dx

1

dt “ ´γ

1

x

1

dx

2

dt “ ´γ

2

x

2

` λ

1

x

1

,

Et x

1

effectue des sauts positifs de loi expo (ν “ e

´x1{b

) ` a taux ϕ “ ϕpx

2

q “ ϕ

1

` ϕ

2

x

2N

1 ` ϕ

3

x

2N

.

§

Voir [Tyran-Kami´ nska et al.

2015] pour l’´ etude en temps long (selon N, ϕ

3

etc...)

§

En cours : ´ etude des

’oscillations’.

(18)

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)

λγ

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 : u

˚

pxq “

Cνpxqγpxq

exp

´ ş

x λpyq

γpyq

dy

¯

(19)

Single cell data on self-regulating gene

Travail en cours (avec O.

Radulescu)

§

Approches

non-param´ etriques (probl` eme mal pos´ e)

§

Approches param´ etriques

(s´ election de mod` eles

r´ egul´ e/non-r´ egul´ e)

(20)

Merci de votre attention !

§

Molecular distributions in gene regulatory dynamics, M.C Mackey, M. Tyran-Kami´ nska and R.Y., Journal of Theoretical Biology (2011) 274 :84-96

§

Dynamic Behavior of Stochastic Gene Expression Models in the Presence of Bursting, M.C Mackey, M. Tyran-Kami´ nska and R.Y., SIAM Journal on Applied Mathematics (2013) 73 :1830-1852

§

Adiabatic reduction of a model of stochastic gene expression

with jump Markov process, R.Y., C. Zhuge, J. Lei, M.C

Mackey, Journal of Mathematical Biology (2014)

68 :1051-1070

Références

Documents relatifs

(a) Demand-driven (the default model in OpenMusic): The user requests the value of node C. This evaluation requires upstream nodes of the graph to evaluate in order to provide C

The current densities obtained here in the “physicochemical” synthetic seawater show the interest of going ahead with the design of a “biological” synthetic seawater, which would

The political and economic interconnections between the choice elements in long term planning are illustrated by the relations between the choice of

We previously defined the original cytotoxic mechanism of anticancerous N-phenyl-N’-(2-chloroethyl)ureas (CEUs) that involves their reactivity towards cellular proteins

Semiconducting SWCNT extraction efficiency plotted as a function of temperature for Pluronic F108 dispersed laser ablation, plasma torch, and arc discharge SWCNTs separated by DGU..

The results of our studies indicate that four factors must be considered when optimizing dye selection and loading to max- imize luminosity of a hybrid SNP dye : molecular

The following discussion will focus on the short to medium term impacts that could be realized within three areas of focus “ feedstock optimization, utilization of microorganisms