Reconstructing neuronal circuitry from spike trains.
Web-application developed by Masahiro Naito, Ryota Kobayashi, and Shigeru Shinomoto (Kyoto University)

This application program estimates inter-neuronal connections from parallel spike trains. For estimation, you may use either [1] Convolutional Neural Network for Estimating synaptic Connectivity from spike Trains (CoNNECT) or [2] Generalized Linear Model applied to Cross Correlation (GLMCC). The references are in the below.

1. prepare a set of spike trians:

Prepare your data in a {.txt} format. The data consists of a set of neuronal spike trains separated by semicolons {;} as


{spike train of the 1st neuron}; 
{spike train of the 2nd neuron};
...
{spike train of the Nth neuron};

Each {spike train} is given as a series of spike times separated by a newline, a comma, or a space. A sample in which spike times are separated by a newline (and spike trains are separated by a semicolon) is shown below


1692.529986
2372.809986
2682.789986
...
;
1396.319986
1405.629986
1713.209986
...
;

In our default setting, spike times should be represented in a unit of [msec], but you can change the setting into [sec] or [µsec].

2. choose an analysis method:
Setting
More details
start
end
min_s
γ
τ
other parameters
bins
WIN
ds
start
end
min_s
γ
τ
other parameters
bins
WIN
ds
start
end
min_s
min_s mode
Setting
Parameters chosen for the experimental CA1 data:
(1) tau=1ms: time scale of interaction kernel was chosen 1ms.
(2) min_s = 1: an interval of ± 1 ms in the cross-correlogram was excluded, because near-synchronous spikes were not detected in this experiment due to the shadowing effect (p6 in the paper).
(3) ds = 1 2 3 4: synaptic delay at each connection was selected automatically.
unit of time
3. estimate connectivity:
0%
 

cross-correlogram: