Challenge D: Predicting Single-Neuron Behavior

Challenge D README V 1.0 2 April 2007

Data from Matthew Larkum (Larkum et al., Cerebral Cortex 14:1059-1070 (2004))

All data in challenge D are from the one single layer 5 pyramidal neuron. In other words, many experiments have been done on the same cell.

For the purpose of the challenge, the data sets are divided into a training and a test data set.
Training data sets contain both the injected current waveforms (e.g. t10S1current.txt, which would be experiment no. 10, current injected at the soma) and the recorded membrane potential (e.g. t10S1.txt, the membrane potential at the soma). In this cell, the membrane potential was always simultaneously recorded at three locations: the soma (S1), a distal dendritic location (D1) and a proximal dendritic location (D2). For example, t10D1.txt would be the membrane potential at the distal dendritic location in experiment no. 10. Note that the recorded membrane potential can be distorted at electrodes which are passing current, but somatic action potentials and the time of their peak can always be detected unequivocally.

The data are given as columns of ASCII floating point numbers. Sampling interval for both currents (pA) and membrane potentials (mV) is 0.1 ms. The first sample in each trace is taken at time 0 ms.

If a current injection waveform at either soma or dendrite is not available, then no current was injected at that location in that experiment.

In the test data sets , where we are withholding the voltage recordings and asking you to predict the somatic action potential times with a precision of 2 ms as in Challenges A and C.

Experiment no. 19 has a slightly different data format. In this experiment, which tests for BAC firing (Larkum et al., Nature 398:338-341 (1999)), each file contains 16 columns of ASCII numbers. They give the membrane potential at the soma (t19S1.txt), distal dendrite (t19D1.txt) and proximal dendrite (t19D2.txt) in response to 16 different current injection protocols as described below. The numbers in the second, fourth, ..., and sixteenth column have been set to zero: these are the trials for which we are asking you to predict both the time course of the membrane potential at the proximal dendritic location (with an accuracy of +- 2 mV), and the times of the somatic action potentials. Since the files t19S1.txt, t19D1.txt and t19D2.txt contain both training and test trials, they are listed in both the training and test directories.

The current injection protocol in experiment no. 19 is a typical BAC firing protocol (Larkum et al. 1999, Fig. 2). At the soma, the same square pulse of current (onset, 15 ms; duration, 5 ms; amplitude, 2.5 nA) is injected in each trial. At the proximal dendritic pipette (D2) zero current is injected. At the distal dendritic pipette (D1) an EPSC-like current waveform is injected (time course, f(t) = (1 - exp(-t/tau1))*exp(-t/tau2) with tau1 = 2 ms and tau2 = 8 ms; peak amplitude of the resulting waveform, 2.6 nA) whose onset is at 9 ms in the first trial and moves by +2 ms in each trial.

For questions, please contact Arnd Roth

Link to Data for Challenge D (version 1.0/April 3)