Table 1.
Geometric parameters for the mouse cone outer segment (COS).
Table 2.
G-protein and effector-related parameters.
Table 3.
Activation and hydrolysis parameters.
Table 4.
Guanylyl cyclase (GC) activity parameters.
αmax is the maximum cGMP synthesis rate in the absence of Ca2+ and αmin is the synthesis rate at saturating Ca2+ concentration. These activities were measured in the absence of bicarbonate.
Table 5.
Parameters for ionic currents of cone outer segments.
Fig 1.
Modeling flash responses of a mouse cone.
Black traces are flash responses recorded from a Gnat1−/− mouse cone for an estimated range of 40—6,000 photoisomerizations with a half-maximal intensity that produced 940 photoisomerizations [45]. The colored traces show model predictions for indicated flash intensities. All simulations use the single set of parameters in Table 10. This set was found by stochastically minimizing the rms error between the model and experimental response solely for the 940-photoisomerizations trace while constraining the parameter values to satisfy known experimental constraints.
Table 6.
Diffusion coefficients for cascade components in the membrane and for second messengers in the cytoplasm.
Table 7.
Strict constraints imposed on many parameters that influence the dark current and the values derived from those parameters.
The range for dark current was centered about the experimental flash response of [45], also shown in Fig 1.
Table 8.
Parameter ranges within which there was no penalty imposed by the Metropolis-Hastings search.
The geometric parameters were held fixed at Rb = 0.6 μm, Rt = 0.4 μm, H = 13.4 μm, ϵ0 = 16.8 nm, ν = 0.65, ω0 = π, and the Ca2+ diffusion coefficient was held constant at DCa2+ = 15 μm2 s−1. [R]σ was omitted since flash response depended only on the initial population of R*, and was otherwise independent of its surface density.
Table 9.
Ranges over which parameters were varied when conducting Sobol sensitivity analysis.
Table 10.
Parameter values for a mouse cone found by minimizing the rms error between experiment and model predictions for a flash producing 940 photoisomerizations according to the Metropolis-Hastings random walk.
Table 11.
Local sensitivity indices for a flash of 940 isomerizations uniformly distributed throughout the outer segment.
Fig 2.
Sobol indices for functionals quantifying E*.
The dot at the center of a circle is the Sobol index obtained by Monte Carlo evaluation (100,00 samples). The blue bars define a 90% confidence interval. Plots show the eight most influential parameters ordered from most significant to least significant. (a) Pairwise sensitivity indices for E* activation. (b) Pairwise sensitivity indices for E* recovery. (c) Pairwise sensitivity indices for peak E* production. (d) Total sensitivity indices for peak E* production.
Fig 3.
Sobol indices for functionals quantifying the drop in current due to flash response, the rms error between simulation and experiment, and the dark current.
Blue bars define a 90% confidence interval (100,000 samples). Plots show the eight most influential parameters. Confidence intervals could be off-center of the estimated Sobol index, because the indices were ratios of two Monte Carlo estimated quantities. (a) Single sensitivity indices for the current drop. (b) Total sensitivity indices for the current drop. (c) Total sensitivity indices for the rms error between model prediction and experiment. (d) Total sensitivity indices for the circulating dark current.
Fig 4.
Sobol indices for functionals quantifying the time-to-peak of the current drop and its overshoot.
Blue bars define a 90% confidence interval (100,000 samples). Plots show the eight most influential parameters. Confidence intervals could be off-center of the estimated Sobol index, because the indices were ratios of two Monte Carlo estimated quantities. (a) Single sensitivity indices for the time-to-peak. (b) Total sensitivity indices for the time-to-peak. (c) Single sensitivity indices for the overshoot. (d) Total sensitivity indices for the overshoot.
Table 12.
As an index approached 1, the functional became dependent only on that parameter. Most values shown are close to 0, which indicated that nonlinear interactions between parameters dominated the cone flash response. While the theoretical value of the Sobol index must fall in the interval [0, 1], small negative values sometimes occurred above as an artifact of the Monte Carlo approximation. These should be regarded as approximately 0. Confidence intervals are given in S1 Appendix.
Table 13.
As an index approached 0, the functional became essentially independent of that parameter. A large index value indicated that the considered parameter contributed to significant nonlinear interactions with other model parameters, so that ignoring it would amount to that index’s loss, as a proportion, of the total variance. Some parameters that were negligible, e.g. mcyc, may have been so because their prescribed uncertainties were smaller than other parameters. Confidence intervals are given in S1 Appendix.