The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Modelling neuronal membranes with Piecewise Deterministic Processes Martin Riedler work with Evelyn Buckwar Heriot–Watt University, Edinburgh, UK Stochastic Models in Neuroscience, CIRM Marseille, Jan 19th , 2010 1 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Outline 1 The Modelling Problem 2 Piecewise Deterministic Processes (PDPs) in finite dimensions 3 Spatial Dynamics – PDPs in infinite dimensions 4 Outlook 2 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Scale levels and modelling approaches I. atomic level II. protein level III. cellular level charged ions involved in large numbers; important is their distribution close to the membrane and their flux across the membrane; flux rate is 103 − 106 ions/ms per open channel; one single channel consists of hundreds of amino acids, hence channel ion and possibly only a few channels involved ∼ low channel density or small fibres; behaviour of transients over length scales of 1µm to 1m; fitting of a deterministic model to the average emergent behaviour; continuous approximation discrete modelling microscopic description discrete modelling hybrid stochastic models discrete particle system macroscopic continuous modelling continuous noise approximation continuous stochastic models deterministic limit macroscopic model ODE/PDE models SDE/SPDE models "adding" noise 3 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Motivation for and aim of our work Motivation several hybrid algorithms proposed that reproduce noisy excitable behaviour for neuron models [Skaugen/Walloe 79, Clay/DeFelice 83, Chow/White 96, etc.]; ”noise = channel noise” in context of neuronal modelling, . . . . . . hardly any theory or rigorous analysis of hybrid processes and their numerical approximation . . . hardly any analysis of continuous stochastic approximations, their appropriateness and quality Our approach and aims use PDPs which provide accurate mathematical description hybrid model (in the space-clamped setting) well known class of stochastic processes with existing theory extend PDP theory to include spatial dynamics develop tools for theoretically analysing hybrid algorithms 4 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Building blocks of a hybrid stochastic model Single channel model – continuous-time Markov chain Example: (simplified) N a-channel → waiting time distr., e.g. P[τ > t] = e−(am (v)+ah (v))t m 0 h0 am (v) bm (v) ah (v) bh (v) m 0 h1 m1 h0 ah (v) bh (v) am (v) bm (v) → post-jump value distr., e.g. am (v) , am (v) + ah (v) ah (v) = . am (v) + ah (v) p m 1 h0 = m1 h1 p m 0 h1 → transmembrane potential constant over time 5 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Building blocks of a hybrid stochastic model Single channel model – continuous-time Markov chain Example: (simplified) N a-channel → waiting time distr., e.g. P[τ > t] = e−(am (v)+ah (v))t m 0 h0 am (v) bm (v) ah (v) bh (v) m 0 h1 → post-jump value distr., e.g. m1 h0 ah (v) bh (v) am (v) bm (v) am (v) , am (v) + ah (v) ah (v) = . am (v) + ah (v) p m 1 h0 = m1 h1 p m 0 h1 → transmembrane potential constant over time (Space-clamped) membrane model – ordinary differential equation C v̇ = m X gi (v) (Ei − v) i=1 conductances gi ∝ fraction of open channels 5 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook combining these building blocks: transmembrane potential is not constant → single channel models are not Markovian anymore the correct waiting distribution for a channel in time-varying potential s 7→ v(s) is, e.g., Example cont. P[τ > t] = e− Rt 0 am (v(s))+ah (v(s))ds 6 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook combining these building blocks: transmembrane potential is not constant → single channel models are not Markovian anymore the correct waiting distribution for a channel in time-varying potential s 7→ v(s) is, e.g., Example cont. P[τ > t] = e− Rt 0 am (v(s))+ah (v(s))ds → using this waiting time distribution [Clay/DeFelice, 83] propose an approximation algorithm for a space-clamped membrane → no analysis of the algorithm; particularly, a strong Markov property of the hybrid process ”(voltage, channel state)” needed for the algorithm to be correct 6 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook (finite-dimensional) Piecewise Deterministic Process - Definition Davis, Markov Models and Optimization, 1994 càdlàg process (y(t))t≥0 , where y(t) = (x(t), r(t)) ∈ Rd × R, with |R| < ∞. continuous, deterministic dynamics – cont. component x(t) a family of ODEs ẋ = gr (x) with solutions t 7→ φr (t; x0 ) for all x(0) = x0 ∈ Rd and every r ∈ R discrete, stochastic dynamics – pwc. component r(t) jump rate λ : Rd × R → R+ , s. t. for the jump times 0 < τ1 < τ2 < . . . of r(t) with τk < T “ Z t ` ´ ” P[τk+1 − T > t|r(T ) = r, x(T ) = x] = exp − λ φr (s; x), r ds 0 d transition measure Q : R × R → P(R) for post-jump values P[r(τk+1 ) = r̄|r(τk+1 −) = r, x(T ) = x] = Q(φr (τk+1 − T ; x), r)({r̄}) 7 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Properties under certain assumptions ... strong, homogeneous Markov process theoretically, an exact simulation algorithm exist; practically, for each inter-jump interval ẋ = gr (x) and, simultaneously, to obtain the next jump time Z t − log U = λ φr (s; x), r ds, 0 with U ∼ U(0,1) , has to be solved numerically. generator A of the Markov process y(t) and its domain D(A) exactly defined, i.e., for f ∈ D(A) Z Af (x, r) = fx (x, r)gr (x) + λ(x, r) [f (x, p) − f (x, r)]Q(x, r)(dp) R → starting point for the approximation with continuous processes generalisation to include spatial dynamics [Buckwar, R., in prep.] 8 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Example: space-clamped, leaky membrane with N channels m 0 h0 am (v) bm (v) ah (v) bh (v) m 0 h1 m1 h0 ah (v) bh (v) am (v) bm (v) m1 h1 v̇ = ḡN a rm1 h1 (EN a − v) − gL v | {z } =gr (v) r = (rm0 h0 , rm1 h0 , rm0 h1 , rm1 h1 ), ri = # channels in state i 9 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Example: space-clamped, leaky membrane with N channels m 0 h0 am (v) bm (v) ah (v) bh (v) m 0 h1 m1 h0 v̇ = ḡN a rm1 h1 (EN a − v) − gL v | {z } =gr (v) ah (v) bh (v) am (v) bm (v) m1 h1 r = (rm0 h0 , rm1 h0 , rm0 h1 , rm1 h1 ), ri = # channels in state i n o X ri = N R = r ∈ R4 : ri ∈ {0 : 1 : N } and i jump rate λ(v, r) = r · am (v)+ah (v), bm (v)+ah (v), am (v)+bh (v), bm (v)+bh (v) transition probabilities, e.g. P[r → r + (−1, 1, 0, 0)] = rm0 h0 am (v) λ(v, r) 9 / 16 The Modelling Problem fin.-dim. PDPs monophasic stimulus # channels =100 150 150 100 100 100 50 50 50 0 0.5 1 0 0 0.5 1 0 150 150 150 100 100 100 50 50 50 0 0 0.5 1 0 0 0.5 Outlook # channels =1000 # channels = 500 150 0 preconditioned stimulus infin.-dim. PDPs 1 0 0 0.5 1 0 0.5 1 10 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook infinite-dimensional Piecewise Deterministic Process - Definition [Buckwar, R., Preprint in prep.] separable, real Hilbert space E continuously and densely embedded in and a Borel subset of another separable, real Hilbert space H; further H ⊂ E ∗ ; càdlàg process (y(t))t≥0 , where y(t) = (u(t), r(t)) ∈ E × R, with |R| < ∞. continuous, deterministic dynamics – cont. component u(t) a family of abstract ODEs u̇ = Lr u + Gr (u) ◦ Lr : E → E ∗ linear operators ◦ Gr : E → E ∗ nonlinear mappings with solutions t 7→ ψr (t; u0 ) for all u(0) = u0 ∈ E and every r ∈ R discrete, stochastic dynamics – pwc. component r(t) jump rate Λ : E × R → R+ , transition measure Q : E × R → P(R) 11 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Properties under certain assumptions ... strong, homogeneous Markov process if the strong (Fréchet-) derivative of f ∈ D(A) w.r.t. the u component exists, then the extended generator of y is given by Z Af (u, r) = hLr u+Gr (u), fu (u, r)i+Λr (u) [f (u, p)−f (u, r)]Qr (dp; u) R with h·, ·i duality pairing of E and fu the unique element of E s.t. df (x, r) ◦ y = (y, fx (x, r))E dx where df du (u, r) ∀y∈E ∈ E ∗ Fréchet derivative of f w.r.t. u at (u, r). → in case of higher (spatial) regularity we can use the inner product on H instead of h·, ·i 12 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Example cont. Ω = [a, b] ⊂ R, E = H01 (Ω), H = L2 (Ω) and C N ∂v(t, x) ∂t = d ∂ 2 v(t, x) X i + gN a (x; r) (EN a − v(t, x)) 4R ∂x2 i=1 v(·, x) = 0, for x = a, b with i −1 gN a (x; r) = ḡN a I{m1 h1 } (ri ) δxi (x) ∈ H (= E ∗ ) ⇒ ∀ v(0) ∈ H01 (Ω) ∃! (weak) solution in C([0, T ], H01 (Ω)). 13 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Example cont. Ω = [a, b] ⊂ R, E = H01 (Ω), H = L2 (Ω) and C N ∂v(t, x) ∂t = d ∂ 2 v(t, x) X i + gN a (x; r) (EN a − v(t, x)) 4R ∂x2 i=1 v(·, x) = 0, for x = a, b with i −1 gN a (x; r) = ḡN a I{m1 h1 } (ri ) δxi (x) ∈ H (= E ∗ ) ⇒ ∀ v(0) ∈ H01 (Ω) ∃! (weak) solution in C([0, T ], H01 (Ω)). R = {m0 h0 , m1 h0 , m0 h1 , m1 h1 }N , N h i X ` ´ am (v(xi )) + ah (v(xi )) Im0 h0 (ri ) + . . . , Λ(v, r) = i P[ri = m0 h0 → ri = m1 h0 ] = am (v(xi )) Λ(v, r) 13 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook 14 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Outlook Conclusion Based on the combination MC model of single channels and the cont. modelling of the charge flow, PDPs are the appropriate description of the resulting stochastic process and provide a reference model. 15 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Outlook Conclusion Based on the combination MC model of single channels and the cont. modelling of the charge flow, PDPs are the appropriate description of the resulting stochastic process and provide a reference model. Work in progress: derive and analyse numerical methods to simulate PDPs use PDP model as reference model for existing hybrid algorithms, convergence and error analysis use PDP model as the starting point to derive and analyse approximations by continuous stochastic models → SDE/SPDE models Further extensions: allow for SDEs/SPDEs instead of the det. inter-jump behaviour other applications, e.g., calcium dynamics 15 / 16 The Modelling Problem fin.-dim. PDPs infin.-dim. PDPs Outlook Thank you for your attention! 16 / 16