encephalogram database

dag.stenberg at nospam.helsinki.fi dag.stenberg at nospam.helsinki.fi
Wed Feb 6 05:09:59 EST 2002

Rich Cooper <richcooper1 at mindspring.com> wrote:
> I'm simply responding to
> the Science News article that seems to indicate there is some kind
> of message being transferred in an impulse other than a simple
> on/off message.  Given that speculation, I'm trying to explore how
> such a message might be encoded and transferred among neurons.

This is a very old idea, going back to times when it was not even known
that there are dozens of different transmitter molecules and hundreds of
receptor molecules with various signal transduction mechanisms. During
the decades that have passed, all this new knowledge about
neurochemistry has emerged, but virtually nothing new yet about possible
coding in the electrical message passed along the axon (waveform or
spike train). 

> I'm suggesting that the molecule transmitted across the synapse
> contains a message encoded in the molecule itself.  Then I'm speculating
> about how the encoding could be carried.  From your comments, you
> seem to have some ideas - how would you think messages might be
> encoded between neurons?

Of course the molecule carries a message, but the impact on the
postsynaptic cell of the same molecule may differ according to the
receptor and the signal transduction pathway. The transmitter molecules
are mostly small, simple molecules that do not change their conformation
(whereas all the postsynaptic components do), as was pointed out here.
What your suggestion must mean is that slightly different amounts of
molecule, or a different combination of molecules can be released
depending on the spike waveform (or to widen the subject, the spike
train time series organization). Frankly, it is difficult to think that
very sophisticated changes in presynaptic axonal electrical waveform 
could be decoded postsynaptically.

> mat (?):
>> Although we say the AP is "all-or-none", the fact is that
>> different action potentials differ in details of amplitude and duration
> and
>> after potential and such depending on the history of firing, the metabolic
>> state of the cell, the chemical environment (ion concentrations), possible
>> activation (phosphorylation) of membrane channels, etc.  And the details
>> of the amplitude and time course of the AP in the presynaptic terminal
>> can be important in determining just how much transmitter is released.
>> Some synaptic modulators work by altering the presynaptic AP in just
>> this way.
> Those are exactly the kinds of considerations I'm referring to.  The real
> AP observed at the output of the neuron depends on a lot of past history,
> and all these differences in one AP versus another could contain the
> message encodings I'm describing.

Quite, but all these presynaptic events converge to the release of the
same simple molecule, and it is difficult to see how any more complex
signal could be contained in this same molecule.

> The original question here was about what frequencies were adequate
> for an encephalogram database.  I'm trying to add to that question the
> possibility that higher frequencies might lead to better understanding
> of the content of the encephalogram.  

I hope you realize that the frequencies in the EEG are not action
potentials, but sets of mainly postsynaptic potential changes, summed
over a huge population of neurons. They work well as a signal that
"something" is happening, but not always well telling us what are the
relevant cellular or molecular events.
  I do think that a moderate measure of knowledge of neurochemistry is a
must even for engineers and mathematicians, who want to add to the
neuroscientific knowledge. Otherwise we get this overemphasis on
analyzing electrical events to dust and beyond.

If we think about an "EEG database", I agree that frequencies up to 60
or 70 Hz should be included. But the amount of data to be sampled and
stored in order to extend the upper frequency even to 250-300 Hz is
formidable. Considering that modeling a single action potential
(although hidden in the EEG in 1000* amplitude-fuls of noise) needs an
upper frequency of at least 20,000 Hz, it does not seem practical to
compose a general database with this whole bandwidth.
> With computer studies, statistical correlations can show useful hints
> about where more depth is needed, even when the mechanism of the
> process isn't understood at the time the correlation is found.  The
> correlation doesn't have to be simple, it just has to be statistically
> significant.

Looking back at the last 40 years of neuroscience, I can testify that
exactly this same argument was put forward in the 1960s, and those who chose
to pursue this line have contributed less than one per cent (if even
that) to the advancement of neuroscience. If my judgement is too harsh,
I would appreciate to be enlightened by the neuroscientists here
present. There may be something that momentarily escapes my attention,
because none of it has seemed relevant for so long.

Dag Stenberg

More information about the Neur-sci mailing list