Feasibility of Estimating Maximum Ion Conductance Parameters from the Shape 
    of the Action Potential. A Simulation Study

Alexandru Dan Corlan; Bogdan Amuzescu; Ivan Milicin; Gheorghe Nistor; Marin Popescu; Adelina Georgescu; Marcel Ovidiu Vlad

Feasibility of Estimating Maximum Ion Conductance Parameters from the Shape of the Action Potential. A Simulation Study

Alexandru Dan Corlan, Bogdan Amuzescu, Ivan Milicin, Gheorghe Nistor, Marin Popescu, Adelina Georgescu, Marcel Ovidiu Vlad,

Biophysical Journal 96:664a-664a, 2009

ABSTRACT

We verified whether the Vm(t) of a myocardial action potential provides sufficient information to characterise the cell regarding ion currents densities. We built a database of 45000 simulated action potentials by running the Luo-Rudy dynamic model with Monte-Carlo generated parameter sets comprising the maximum conductances for 12 major current components. Each parameter was generated uniformly in the range of 0.5-2.0 times the default. We then simulated experiments in which a 'data' action potential (V_d(t)) was generated randomly in the same parameter interval and using the same pacing protocol and we tried to estimate its parameters. Each AP in the database was assigned the same prior probability. Then at each of 50 steps spaced at 5 ms a posterior probability was computed that was used as the prior probabily for the next step. We considered for each step t(j) the difference DV(ij) between the i'th action potential in the database and the 'data' action potential. Using a normal noise model we calculated the non-conditional probability of DV(ij) assuming it has a normal distribution then the post-probability for step j given the prior obtained in step j-1. We took the parameters of the database element with the maximum posterior probability after the last step as the best estimate of the parameters for the 'data' action potential. The absolute difference between estimated and data parameters was used to assess the quality of estimation. RMSD the root mean square difference between the V(t) reconstructed from the estimated parameters and V_d(t) was used to measure the quality of fit. In 100 experiments as described above we found a RMSD of 4.14 +/- 1.15 mV (mean +/- SD) corresponding to a very close resemblance of the data to the estimated V(t). However the absolute differences in parameters were large ranging from 0.30 +/- 0.31 for I_Kr to 0.9 +/- 0.5 for I_Na.

[Publisher]