Central Synapse — Voltage-Clamp EPSC Simulator

Each stimulus is delivered to 3 independent release sites, each releasing one vesicle with probability P_r after a ~1 ms synaptic delay (presynaptic Ca²⁺ channel opening, vesicle fusion, diffusion across the cleft, and receptor binding — Sabatini & Regehr, 1996). Released glutamate activates both AMPA and NMDA receptors on the same postsynaptic density: the AMPA current is fast (rise ~0.3 ms, decay ~3 ms) and the NMDA current is slow (rise ~5 ms, decay ~80 ms) and voltage-dependent, since extracellular Mg²⁺ blocks the NMDA channel pore at hyperpolarized potentials and unblocks as the membrane depolarizes (Jahr & Stevens relation). The recorded current is I(t) = g(t) · (Vhold − 0 mV), so at very negative holding potentials the trace deflects sharply downward (inward current), and near 0 mV the driving force — and the current — shrinks toward zero even though channels are open.

Unlike the neuromuscular junction's single terminal with hundreds of release sites, a central synaptic connection typically involves only a handful of contacts, so failures (0 quanta released) are common even without lowering calcium — this is the everyday operating regime of a cortical or hippocampal synapse, not a special experimental condition.

In the Paired-Pulse / Train tab, each release site depletes after releasing and recovers over hundreds of milliseconds, while each release event also transiently raises release probability at the remaining sites (facilitation). At low P_r, facilitation dominates the paired-pulse ratio; at higher P_r, depletion-driven depression takes over — the same rule of thumb described by Dobrunz & Stevens (1997).

Because a connection only has 2–4 release sites, the paired-pulse ratio itself is a noisy statistic: with so few sites, whether a given site happens to still be available for pulse 2 is close to a coin flip, and averaging a modest number of trials doesn't fully wash that out — a fresh "average" run can genuinely land on facilitation one time and depression the next, purely from sampling variability, even with every slider held fixed. This is a real feature of small-N synapses (and part of why single-connection experiments need many repeats), not a simulator error. Averaging more trials narrows the spread but only slowly, since it shrinks with the square root of trial count.