References

1

397.5 UU84
Methods

• Recordings
• Electrophysiological recordings were conducted in
  the hippocampus and mPFC of male
  isoflurane-anaesthetised Lister-hooded rats,
  weighing 200-350g. Multiple single-unit and LFP
  activity were recorded with a Plexon MAP system
  using micro-electrode arrays placed in each
  structure. Recording sites were confirmed by
  histology.
• Rats were administered KA (10 mg kg-1) and
  vehicle (5 EtOH / 5 Cremophor EL (Fluka
  Biochemika) / 90 saline v/v n4). After a 33
  min period of basal recording, KA was
  administered intra-peritoneally resulting in a
  period of approximately 20-30 min to onset of
  drug-evoked responses 2,3.
• Two recording sets are
• presented and analysed here
• Cases 1 and 2.

• LFPs Numerical Implementation
• Spike data processing, MAR/PDC calculation and
  plotting were performed with our MATLAB software
  package which includes scripts from the
  toolboxes Biosig (http//biosig.sourceforge.net
  ) and NeuroSpec (http//www.neurospec.org).

• Model order selection criteria (Bayesian
  Information Criterion (BIC), Akaike Criterion)
  failed to yield an optimal model order (p). They
  kept decreasing with increasing model order.

• Representation of the most medial (?) and
  lateral (?) dorsal hippocampal recording sites.
• Representation of the prelimbic mPFC (?)
  recording sites.
• Representative trace of basal activity in LFPs
  recorded from mPFC, medial hippocampus and
  lateral hippocampus.

• We set the model order to p 70. It yields a
  MAR model that produces the original LFP power
  spectrum accurately. Much lower p gives models
  with bad fit. Much higher p does not particularly
  improve the fit.

• Spike Trains Firing Rates processing
• Each spike train is divided into (overlapping)
  bins. The mean firing rate is calculated in each
  bin.
• Principal Components Analysis (PCA) is performed
  to group the firing rates of each brain region
  together.
• The first principal component (PC1) of each group
  is kept so as to end up with one continuous
  function for each brain region.

• Spike Trains Spike Times processing
• We apply the French-Holden Algorithm 4 with
  Nyquist frequency fN.
• Each spike (ti) in a spike train is convolved
  with the kernel
• Equal to filtering with a low-pass
• phase-less filter with cut-off
• frequency fN.
• The resulting continuous function
• is sampled with sampling rate

spikes

• References
• Baccalá L.A., Sameshima K., 2001, Biol. Cybern,
  84 463-474.
• Coomber B., O'Donoghue M. F., Mason R., 2008,
  Synapse, 62 746-755.
• Coomber B., Taxidis I., Owen M., Mason R., 2008,
  in preparation.
• French A.S., Holden A.V., 1971, Biol. Cyb., 8
  165-171.
• Granger C.W.J., 1969, Econometrica, 37 424-438.
• Hammond C., 2001, Academic Press.
• Isomura Y. et al., 2006, Neuron, 52 871-882.
• Jay T.M., Witter M.P., 1991, J. Comp. Neurol.,
  313 574-586.
• Lütkepohl H., 1993, Springer.

Acknowledgments This research was supported by
the Medical Research Council, University of
Nottingham and by a Marie Curie Research Training
Fellowship
Lab web site www.nottingham.ac.uk/neuronal-netwo
rks