Quantifying synaptic efficacy

Abstract

time. In such a condition, there is a large degree of uncertainty about whether the next symbol in the string will be a '0' or a '1'. But if we also know what a given presynaptic input is doing during the same period, this uncertainty will decrease. So, if the presynaptic input is strong, the prob-ability of obtaining a '1' as post-synaptic output will increase. In the extreme case, a very strong input will always make the postsynaptic neuron fire, whereas a very weak input will have no effect. The authors devel-oped a compression algorithm that allowed them to use their knowledge of the input to simplify the post-synaptic output string. So, for a very strong input, we can predict with confidence what the postsynaptic output will be, and therefore com-press the string that we need to rep-resent it. The SIE is a measurement of how many bits of information were saved by this compression. By measuring SIE while altering the amplitude, kinetics and dendritic location of the presynaptic input, the authors obtained quantitative esti-mates of how information efficacy changed as a function of different synaptic properties. To make their model more realistic, they measured SIE in the presence of a high level of background synaptic activity, and used increasingly complex models of the postsynaptic neuron. Finally, London et al. measured SIE experi-mentally, and found that some pre-dictions of their computational analysis could be confirmed in real neurons. Critics of theoretical approaches might argue that this type of analysis is necessarily limited by our under-standing of the cellular and biophysi-cal characteristics of actual synapses and neurons. However, as the models incorporate more features of real cells, the insights that we will obtain from SIE as a measure of information efficacy should continue to increase. So, the quantitative nature of SIE provides us with a useful tool to answer the apparently simple ques-tion of what a single synapse tells the postsynaptic axon.