Fig. 6: Phase-to-rate recoding improves STDP in CA3.
a Schematic of the simple CA3 model and its input. b Representative examples of DG spike trains (left) and mean GC firing rates (right), comparing the full and the GC-rate-controlled no-feedback (no fb) pydentate model. c Representative examples of the spike trains and mean firing rates elicited by this input in samples of CA3 pyramidal cells (top) and CA3 feedforward interneurons (bottom). d Pyramidal and interneuron rates in response to varying the inhibitory interneuron output (mean ± sd). e Illustration of symmetric STDP kernel leading to synaptic potentiation independent of spike sequence (arb. units indicates arbitrary units). f Mean synaptic potentiation after 2 s simulation for inhibition=2 mV and tauSTDP = 20 ms. g Sensitivity analysis of mean synaptic potentiation for variations in inhibitory strength (as in d) and tauSTDP for the full (left) and no-feedback (middle) networks, as well as their ratio (right). h Left, mean synaptic potentiation (f) normalized to the equivalent ratio of pyramidal cell firing rates (as in c, d). Right, sensitivity analysis for normalized synaptic potentiation. i–l Same as e–h, but for asymmetric STDP. Asterisks indicate significance in two-sided t tests. Asterisks indicate significance at p < 0.0001 in unpaired two-tailed t test with Welch’s correction; n = 30, 23 grid seeds for full and no feedback, respectively. Box plots show the median, the interquartile range (box) and the data range w/o Tukey-outliers (whiskers). Source data are provided in Source data.xlsx. Full statistics are shown in Supplementary Tables 13 and 14.
