HAL Id: hal-02797058
https://hal.inrae.fr/hal-02797058
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.
Nucleation time in stochastic Becker-Doring model
Romain Yvinec, Samuel Bernard, Tom Chou, Julien Deschamps, Maria R.
d’Orsogna, Erwan Hingant, Laurent Pujot-Menjouet
To cite this version:
Nucleation Time in Stochastic Becker-D¨
oring
Model
Romain Yvinec1, Samuel Bernard2,3, Tom Chou4, Julien
Deschamps5, Maria R. D’Orsogna6, Erwan Hingant7 and
Laurent Pujo-Menjouet2,3
1
BIOS group INRA Tours, France.
2Institut Camille Jordan, Universit´e Lyon 1, Villeurbanne, France. 3
INRIA Team Dracula, Inria Center Grenoble Rhˆone-Alpes, France. 4Depts. of Biomathematics and Mathematics, UCLA, Los Angeles, USA.
5
DIMA, Universita degli Studi di Genova, Italy. 6Dept. of Mathematics, CSUN, Los Angeles, USA. 7
Numerical results
Coarse-graining
Outline
Amyloid diseases and Becker-D¨oring model
Protein accumulation in amyloid by
nucleation-polymerization
Protein accumulation in amyloid by
nucleation-polymerization
Times series of in-vitro spontaneous polymerization
Xue et al. PNAS (2008)
Quantification of experiment
ThT fluorescence « Number of polymerized protein
Becker-D¨
oring model
Becker-D¨
oring model
Reversible one-step agregation Ci ` C1 pi ÝÝá âÝÝ qi `1 Ci `1 (1)
§ Purely dynamic model (law of mass-action) : no space, no polymer
Becker-D¨
oring model
Reversible one-step agregation Ci ` C1 pi ÝÝá âÝÝ qi `1 Ci `1 (1)
§ Indirect interaction between polymer Ci, i ě 2 via the available
number of monomers C1.
C1ptq `
ÿ
i ě2
Becker-D¨
oring model
Reversible one-step agregation Ci ` C1 pi ÝÝá âÝÝ qi `1 Ci `1 (1)
§ In spontaneous polymerization experiment,
§ Initial condition given by Cipt “ 0q “ 0 @i ě 2. § Measured variable :ř
Becker-D¨
oring model
Reversible one-step agregation Ci ` C1 pi ÝÝá âÝÝ qi `1 Ci `1 (1)
§ The (observed) nucleation time is given by
inftt ě 0 : ÿ
i ěN
iCiptq ě ρM | Cipt “ 0q “ Mδi “1u .
Becker-D¨
oring model
Reversible one-step agregation Ci ` C1 pi ÝÝá âÝÝ qi `1 Ci `1
§ What are the dependencies of the nucleation time with
respect to the model parameters ?
total mass : M ; nucleus size : N
aggregation rates : pi, i ě 1 fragmentation rates : qi, i ě 2
§ What is the nucleation time for very large initial quantity M
and nucleus size N ?
§ In experiment, M « 1010´ 1015,
Deterministic BD model and Classical Nucleation Theory
$ ’ ’ ’ & ’ ’ ’ % dci dt “ Ji ´1´ Ji, i ě 2 , Ji “ pic1ci´ qi `1ci `1, i ě 1 , dc1 dt “ ´J1´ ř8 i “1Ji.Equilibrium is given by Ji ” J “ 0, which implies
ci “ Qic1i , Qi “
p1p2¨ ¨ ¨ pi ´1
q2q3¨ ¨ ¨ qi
The mass at equilibrium is given by ρpc1q “
ÿ
i ě1
iQic1i
If this series has a finite radius of convergence, zs, then there is a
critical mass
ρs “
ÿ
i ě1
Deterministic BD model and Classical Nucleation Theory
$ ’ ’ ’ & ’ ’ ’ % dci dt “ Ji ´1´ Ji, i ě 2 , Ji “ pic1ci´ qi `1ci `1, i ě 1 , dc1 dt “ ´J1´ ř8 i “1Ji.[Ball, Carr, Penrose, CMP (1986)]
If M ď ρs, then (with strong convergence)
lim
tÑ8ciptq “ Qiz
i, ρpzq “ M
If M ą ρs, then (with weak convergence)
lim
tÑ8ciptq “ Qiz i
s, M ´ ρpzsq ““loss of mass”
As M Œ ρs, there is a solution for which Ji « J˚ is exponentially
small, and
Deterministic BD model – Some remarks
§ For constant or linear kinetic rates pi, qi, one can reduce the
system to 1 or 2 ODEs on ÿ i ě1 ci, ÿ i ě1 ici.
§ Based on scaling arguments, one can show that for qi “ 0
(irreversible nucleation),
inftt ě 0 : cNptq ě ρM | cipt “ 0q “ Mδi “1u »
1
M .
while for “qi Ñ 82 (pre-equilibrium nucleation),
inftt ě 0 : cNptq ě ρM | cipt “ 0q “ Mδi “1u »
Stochastic Becker-D¨
oring model
Reversible one-step agregation Ci ` C1 pi ÝÝá âÝÝ qi `1 Ci `1
We define a continuous time Markov Chain on
tpCiqi ě 1 : Ci P N,
ÿ
i ě1
iCi “ Mu
Transitions are given by
Stochastic Becker-D¨
oring model
Reversible one-step agregation Ci ` C1 pi ÝÝá âÝÝ qi `1 Ci `1
Outline
Amyloid diseases and Becker-D¨oring model
Numerical results
§ Law of large numbers as
M Ñ 8 [Jeon. CMP (1998)]
§ Any macroscopic quantity like.
inftt ě 0 : ÿ
i ěN
iCiptq ě ρM
| Cipt “ 0q “ Mδi “1u .
converges (if reachable) to a finite deterministic value as M Ñ 8.
§ This may not be true for
microscopic quantity, for instance.
inftt ě 0 : CNptq ě 1
| Cipt “ 0q “ Mδi “1u .
[Y., D’Orsogna, Chou JCP (2012)]
§ Non-monotonous w.r.t
reaction rate
§ Bimodal for ’small’
fragmentation rate
Outline
Amyloid diseases and Becker-D¨oring model
Numerical results
Coarse-graining
When N Ñ 8?
We start from a rescaled model (ε “ 1{N, ε2“ 1{M)
$ ’ & ’ % dciε dt “ 1 ε“J ε i ´1´ Jiε‰ , i ě 2 , mε “ c1εptq ` ε2ÿ i ě2 iciεptq . With fεpt, x q “ři ě2ciεptq1rpi ´1{2qε,pi `1{2qεqpx q,
Scaling idea : excess of monomer c1εptq “ ε2c1ptq ,
Compensated aggregation / fragmentation pεpx q “ ÿ i ě2 ppi{ε2q1rεi ,εpi `1qr qεpx q “ ÿ i ě3 qi1rεi ,εpi `1qr
and slow first step :
When N Ñ 8?
We start from a rescaled model (ε “ 1{N, ε2“ 1{M)
$ ’ & ’ % dciε dt “ 1 ε“J ε i ´1´ Jiε‰ , i ě 2 , mε “ c1εptq ` ε2ÿ i ě2 iciεptq . With fεpt, x q “ři ě2ciεptq1rpi ´1{2qε,pi `1{2qεqpx q,
From the polymer point of view, we have accelerated fluxes, all of the same order :
When N Ñ 8?
We start from a rescaled model (ε “ 1{N, ε2“ 1{M)
$ ’ & ’ % dciε dt “ 1 ε“J ε i ´1´ Jiε‰ , i ě 2 , mε “ c1εptq ` ε2ÿ i ě2 iciεptq . With fεpt, x q “ři ě2ciεptq1rpi ´1{2qε,pi `1{2qεqpx q,
d dt ż`8 0 fεpt, x qϕpx q dx ““p1εc1εptq2´ q2c2εptq ‰ ˜ 1 ε ż5{2ε 3{2ε ϕpxq dx ¸ ` ż`8 0 rpεpx qc1εptqfεpt, x q∆εϕpxq ´ qεpx qfεpt, x q∆´εϕpxqs dx ,
Theorem (Deschamps, Hingant, Y. (2016))
we have fε Ñ f (weakly in X “ tν P Mbpr0, 8q : ş xνpdxq ă 8u)
solution of d dt ż`8 0 f pt, xqϕpxq dx “ ż`8 0 rppx qc1ptq ´ qpx qs ϕ1px qf pt, x q dx ,
and c1ptq `ş xf pt, xq “ m, for all ϕ P Ccp0, 8q.
This is the weak form of Bf
Bt `
BpJpx , tqf pt, x qq
d dt ż`8 0 fεpt, x qϕpx q dx ““p1εc1εptq2´ q2c2εptq ‰ ˜ 1 ε ż5{2ε 3{2ε ϕpxq dx ¸ ` ż`8 0 rpεpx qc1εptqfεpt, x q∆εϕpxq ´ qεpx qfεpt, x q∆´εϕpxqs dx ,
Theorem (Deschamps, Hingant, Y. (2016)) When c1p0q ą limx Ñ0 qpxqppxq, d dt ż`8 0 f pt, xqϕpxq dx “ Nptqϕp0q ` ż`8 0 rppx qc1ptq ´ qpx qs ϕ1px qf pt, x q dx ,
for all ϕ P Cbp0, 8q, which gives the boundary condition
lim
d dt ż`8 0 fεpt, x qϕpx q dx ““p1εc1εptq2´ q2c2εptq ‰ ˜ 1 ε ż5{2ε 3{2ε ϕpxq dx ¸ ` ż`8 0 rpεpx qc1εptqfεpt, x q∆εϕpxq ´ qεpx qfεpt, x q∆´εϕpxqs dx ,
Theorem (Deschamps, Hingant, Y. (2016))
Nptq is an explicit function of c1ptq, and is given by a quasi
steady-state approximation of c2ε“ fεpt, 2εq, given by the solution of #
0 “ rJi ´1pc1q ´ Jipc1qs , i ě 2 ,
c1ptq “ c1.
When c1 ą limx Ñ0qpxqppxq, the solution of Ji ” J ‰ 0 is linked to the
Large nucleus N „
?
M
§ First case (pp0qm ą qp0q) : Convergence towards a
deterministic value.
Large nucleus N „
?
M
§ Second case (pp0qM ă qp0q) : Exponentially large time and
Outline
Amyloid diseases and Becker-D¨oring model
Numerical results Coarse-graining
Quantifying the large deviation in toy model
A much simpler version of this model consider that a single aggregate may be formed at a time :
k ÝÝÝÝÝÝáâÝÝÝÝÝÝpkpm´kεq
qk`1
k ` 1 , which converges (with time rescaling) to
dx
Quantifying the large deviation in toy model
A much simpler version of this model consider that a single aggregate may be formed at a time :
k ÝÝÝÝÝÝáâÝÝÝÝÝÝpkpm´kεq
qk`1
k ` 1 , which converges (with time rescaling) to
dx
dt “ ppx qpm ´ x q ´ qpx q
§ To leading order the stationary
prob. density is u˚ px q “ Ce ´1εşxlog ´ qpy q ppy qpm´y q ¯ dy a ppxqpm ´ xqqpxq . § MFPT is explicit and is exponentially large in ε
Quantifying the large deviation in toy model
Can we perform similar calculations with n clusters ?
pk0, k1q pk0pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk0`1 pk0` 1, k1q , pk0, k1q pk1pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk1`1 pk0 , k1` 1q ,
which converges (with time rescaling) to
dx
dt “ ppxqpm ´ x ´ y q ´ qpxq
dy
Quantifying the large deviation in toy model
Can we perform similar calculations with n clusters ?
pk0, k1q pk0pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk0`1 pk0` 1, k1q , pk0, k1q pk1pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk1`1 pk0 , k1` 1q ,
which converges (with time rescaling) to
dx
dt “ ppxqpm ´ x ´ y q ´ qpxq
dy
Quantifying the large deviation in toy model
Can we perform similar calculations with n clusters ?
pk0, k1q pk0pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk0`1 pk0` 1, k1q , pk0, k1q pk1pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk1`1 pk0 , k1` 1q ,
which converges (with time rescaling) to
dx
dt “ ppxqpm ´ x ´ y q ´ qpxq
dy
Quantifying the large deviation in toy model
Can we perform similar calculations with n clusters ?
pk0, k1q pk0pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk0`1 pk0` 1, k1q , pk0, k1q pk1pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk1`1 pk0 , k1` 1q ,
which converges (with time rescaling) to
dx
dt “ ppxqpm ´ x ´ y q ´ qpxq
dy
Quantifying the large deviation in toy model
Can we perform similar calculations with n clusters ?
pk0, k1q pk0pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk0`1 pk0` 1, k1q , pk0, k1q pk1pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk1`1 pk0 , k1` 1q ,
which converges (with time rescaling) to
dx
dt “ ppxqpm ´ x ´ y q ´ qpxq
dy
Quantifying the large deviation in toy model
Can we perform similar calculations with n clusters ?
pk0, k1q pk0pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk0`1 pk0` 1, k1q , pk0, k1q pk1pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk1`1 pk0 , k1` 1q ,
which converges (with time rescaling) to
dx
dt “ ppxqpm ´ x ´ y q ´ qpxq
dy
Quantifying the large deviation in toy model
Can we perform similar calculations with n clusters ?
pk0, k1q pk0pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk0`1 pk0` 1, k1q , pk0, k1q pk1pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk1`1 pk0 , k1` 1q ,
which converges (with time rescaling) to
dx
dt “ ppxqpm ´ x ´ y q ´ qpxq
dy
Quantifying the large deviation in toy model
Can we perform similar calculations with n clusters ?
pk0, k1q pk0pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk0`1 pk0` 1, k1q , pk0, k1q pk1pm´pk0`k1qεq ÝÝÝÝÝÝÝÝÝÝá âÝÝÝÝÝÝÝÝÝÝ qk1`1 pk0 , k1` 1q ,
which converges (with time rescaling) to
dx
dt “ ppxqpm ´ x ´ y q ´ qpxq
dy
Next :
§ Proving Large Deviation Principle for the full (S)BD
§ Quantifying the MFPT
§ Data fitting in spontaneous polymerization experiment