Information theory-based approach towards studying anti-coincidence detection via graded amplitude dendritic action potentials.

In contrast to typical all or none action potential, recent discovery of graded amplitude action potentials in cortical neurons enabled the dendrites to perform XOR computation, previously thought to be performed only at network level. Thus, these special neurons can perform anti-coincidence detection at the dendritic level, but a lot is unanswered about this phenomenon. Can such experimentally observed dendritic action potential generating system transmit information about stimuli having varying degrees of temporal overlap? Can the system add to the repertoire of computations performed at dendritic level by enhancing the information transmission about varying amplitude stimuli? In this information theory-based study done in single compartment and two-compartment dendritic models, it is shown that such a system can indeed transmit information about the temporal overlap of stimuli as well as amplitudes of stimuli even at high input noise levels. First, the calculation of mutual information between single stimulus and response i.e. I(S;R) with varying noise showed that the information about temporally overlapping nature of stimuli is precisely transmitted by such a system. Secondly, the time evolution of mutual information was simulated through data from the system and it positively reinforced the above-mentioned result. Next, varying amplitude input stimuli was provided to the system and calculation of mutual information between two stimuli and one response i.e. I(S1,S2;R) with varying noise levels revealed that such a system optimally transmits the information about stimuli even at high noise levels. Finally, calculation of this information measurement with respect to time in an experiment with constant overlap but varying input amplitude again positively reinforced the result. Key PointsO_LIInformation theory-based measurements were employed to assess the role of graded amplitude dendritic action potentials. C_LIO_LIAction potentials (APs) with maximal amplitudes for threshold level stimuli and lower amplitudes for stronger stimuli were modelled with high voltage Ca2+ (HVA like) channels, Ca2+ dependent (BK-like) channel, leak channel and calcium pump in a single compartment model and two compartment dendritic model. C_LIO_LIAnalysis done here, on comparison with control compartment generating constant amplitude AP via standard Hodgkin-Huxley sodium potassium channel revealed that such a compartment shows optimal information transmission about both varying amplitudes input current stimuli as well as varying time overlap stimuli. C_LI