Limits...
Probabilistic inference of short-term synaptic plasticity in neocortical microcircuits.

Costa RP, Sjöström PJ, van Rossum MC - Front Comput Neurosci (2013)

Bottom Line: We demonstrate that for typical synaptic dynamics such fitting may give unreliable results.Using our result we propose a experimental protocol to more accurately determine synaptic dynamics parameters.Our approach to demarcate connection-specific synaptic dynamics is an important improvement on the state of the art and reveals novel features from existing data.

View Article: PubMed Central - PubMed

Affiliation: Institute for Adaptive and Neural Computation, School of Informatics, University of Edinburgh Edinburgh, UK.

ABSTRACT
Short-term synaptic plasticity is highly diverse across brain area, cortical layer, cell type, and developmental stage. Since short-term plasticity (STP) strongly shapes neural dynamics, this diversity suggests a specific and essential role in neural information processing. Therefore, a correct characterization of short-term synaptic plasticity is an important step towards understanding and modeling neural systems. Phenomenological models have been developed, but they are usually fitted to experimental data using least-mean-square methods. We demonstrate that for typical synaptic dynamics such fitting may give unreliable results. As a solution, we introduce a Bayesian formulation, which yields the posterior distribution over the model parameters given the data. First, we show that common STP protocols yield broad distributions over some model parameters. Using our result we propose a experimental protocol to more accurately determine synaptic dynamics parameters. Next, we infer the model parameters using experimental data from three different neocortical excitatory connection types. This reveals connection-specific distributions, which we use to classify synaptic dynamics. Our approach to demarcate connection-specific synaptic dynamics is an important improvement on the state of the art and reveals novel features from existing data.

No MeSH data available.


The performance of various stimulation protocols to infer short-term plasticity parameters. (A) Comparison of mean sample estimation error using different stimulation trains. Black arrow corresponds to the protocol used in Figure 2. A periodic spike train at 30 Hz with eight pulses and nine recovery pulses [green arrow, (B)] already provided a better estimate of the STP parameters. However, a Poisson train provided a much smaller error when using more than 20 spikes with a close to zero error for 1000 spikes [blue arrow, (C)]. (B) Posterior distribution for a periodic train with nine recovery pulses (cf. Figure 2B). (C) Posterior distribution for a Poisson train with 1000 pulses. The true parameters are shown as filled circles and the MAP solutions as diamonds. For visualization the marginal probabilities were scaled by their standard deviation.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3674479&req=5

Figure 3: The performance of various stimulation protocols to infer short-term plasticity parameters. (A) Comparison of mean sample estimation error using different stimulation trains. Black arrow corresponds to the protocol used in Figure 2. A periodic spike train at 30 Hz with eight pulses and nine recovery pulses [green arrow, (B)] already provided a better estimate of the STP parameters. However, a Poisson train provided a much smaller error when using more than 20 spikes with a close to zero error for 1000 spikes [blue arrow, (C)]. (B) Posterior distribution for a periodic train with nine recovery pulses (cf. Figure 2B). (C) Posterior distribution for a Poisson train with 1000 pulses. The true parameters are shown as filled circles and the MAP solutions as diamonds. For visualization the marginal probabilities were scaled by their standard deviation.

Mentions: To quantify which protocol allows for the most precise recovery of the true parameters of simulated STP data (Figure 3A), we computed the sample estimation error over N = 22,500 MCMC samples to the true parameters , as , where the average is over all the runs and all five parameter sets (Table 1). To achieve similar weighting, the parameters were normalized to the true parameters. Alternatively, we normalized the estimated parameters on the upper limit of their priors, or we omitted normalization altogether. This yielded similar results. Note that in probabilistic spirit, this error also quantifies the spread in the distribution. A smaller E gives more peaked distributions, which correspond to tighter parameter estimates. Note that, although similar, this error measure does not follow the standard bootstrap approach.


Probabilistic inference of short-term synaptic plasticity in neocortical microcircuits.

Costa RP, Sjöström PJ, van Rossum MC - Front Comput Neurosci (2013)

The performance of various stimulation protocols to infer short-term plasticity parameters. (A) Comparison of mean sample estimation error using different stimulation trains. Black arrow corresponds to the protocol used in Figure 2. A periodic spike train at 30 Hz with eight pulses and nine recovery pulses [green arrow, (B)] already provided a better estimate of the STP parameters. However, a Poisson train provided a much smaller error when using more than 20 spikes with a close to zero error for 1000 spikes [blue arrow, (C)]. (B) Posterior distribution for a periodic train with nine recovery pulses (cf. Figure 2B). (C) Posterior distribution for a Poisson train with 1000 pulses. The true parameters are shown as filled circles and the MAP solutions as diamonds. For visualization the marginal probabilities were scaled by their standard deviation.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3674479&req=5

Figure 3: The performance of various stimulation protocols to infer short-term plasticity parameters. (A) Comparison of mean sample estimation error using different stimulation trains. Black arrow corresponds to the protocol used in Figure 2. A periodic spike train at 30 Hz with eight pulses and nine recovery pulses [green arrow, (B)] already provided a better estimate of the STP parameters. However, a Poisson train provided a much smaller error when using more than 20 spikes with a close to zero error for 1000 spikes [blue arrow, (C)]. (B) Posterior distribution for a periodic train with nine recovery pulses (cf. Figure 2B). (C) Posterior distribution for a Poisson train with 1000 pulses. The true parameters are shown as filled circles and the MAP solutions as diamonds. For visualization the marginal probabilities were scaled by their standard deviation.
Mentions: To quantify which protocol allows for the most precise recovery of the true parameters of simulated STP data (Figure 3A), we computed the sample estimation error over N = 22,500 MCMC samples to the true parameters , as , where the average is over all the runs and all five parameter sets (Table 1). To achieve similar weighting, the parameters were normalized to the true parameters. Alternatively, we normalized the estimated parameters on the upper limit of their priors, or we omitted normalization altogether. This yielded similar results. Note that in probabilistic spirit, this error also quantifies the spread in the distribution. A smaller E gives more peaked distributions, which correspond to tighter parameter estimates. Note that, although similar, this error measure does not follow the standard bootstrap approach.

Bottom Line: We demonstrate that for typical synaptic dynamics such fitting may give unreliable results.Using our result we propose a experimental protocol to more accurately determine synaptic dynamics parameters.Our approach to demarcate connection-specific synaptic dynamics is an important improvement on the state of the art and reveals novel features from existing data.

View Article: PubMed Central - PubMed

Affiliation: Institute for Adaptive and Neural Computation, School of Informatics, University of Edinburgh Edinburgh, UK.

ABSTRACT
Short-term synaptic plasticity is highly diverse across brain area, cortical layer, cell type, and developmental stage. Since short-term plasticity (STP) strongly shapes neural dynamics, this diversity suggests a specific and essential role in neural information processing. Therefore, a correct characterization of short-term synaptic plasticity is an important step towards understanding and modeling neural systems. Phenomenological models have been developed, but they are usually fitted to experimental data using least-mean-square methods. We demonstrate that for typical synaptic dynamics such fitting may give unreliable results. As a solution, we introduce a Bayesian formulation, which yields the posterior distribution over the model parameters given the data. First, we show that common STP protocols yield broad distributions over some model parameters. Using our result we propose a experimental protocol to more accurately determine synaptic dynamics parameters. Next, we infer the model parameters using experimental data from three different neocortical excitatory connection types. This reveals connection-specific distributions, which we use to classify synaptic dynamics. Our approach to demarcate connection-specific synaptic dynamics is an important improvement on the state of the art and reveals novel features from existing data.

No MeSH data available.