Re: Signal Processing (was BLUE GENE)

From: Martin Striz (mstriz@gmail.com)
Date: Wed Apr 12 2006 - 10:21:29 MDT


On 4/12/06, Russell Wallace <russell.wallace@gmail.com> wrote:

> > Why propagate a signal as a transiet voltage change rather than a
> > direct electric (or electrochemical) current? Why use absurdly
> > complicated intracellular signal transduction pathways? Why implement
> > over 100 neurotransmitters when one will do just fine (modular
> > specificity can be achieved through wiring or through postsynaptic
> > receptor response)?
> >

> Hmm, don't know enough about the second to comment at all... on the third,
> why would the alternatives be better than using extra neurotransmitters?

Simplicity, efficiency, energy conservation. Since you have to build
a different receptor for each neurotransmitter, why not just build
different receptors that respond to the same molecule but transduce
the signal differently. >100 neurotransmitters, at least 80 of which
are non-peptides, require their own biosynthetic pathways (although
some of them overlap, such as tyrosine -> L-dopa -> dopamine ->
epinephrine -> norepinephrine, the last three of which are all used as
neurotransmitters, but most of them are independent pathways).
Assuming that there are on average three enzymes necessary to get from
a precursor to a finished neurotransmitter product, that's an extra
300 genes which take up cellular resources. By no coincidence, about
half of the genes in the genome are expressed in the brain, more than
in any other organ.

> The first is an interesting question... neural signals appear inefficient
> in terms of raw speed, but they are extraordinarily efficient in terms of
> energy expended per bit sent, far more so than anything we can build today.

Is that true? I don't know. According to a number of sources
(http://hypertextbook.com/facts/2001/JacquelineLing.shtml), the brain
consumes about 23 J/s. In other words, it's a 23 W processor with a
capacity of about 10^15 to 10^18 computer-equivalent flops. How much
power does a typical microchip draw?

> Is that inherent to the method of transmission? Or is there a way to be both
> fast and energy efficient with direct electric current signals given
> sufficiently advanced nanotechnology?

The savings is probably due to efficiency. Chips expend a lot of heat
because they use resistance for processing. That's just wasted
energy, aka cost. Your head doesn't heat up to 80C (or whatever) when
you think. If nanotechnology continues to use resistance, don't
expect it to be as efficient.

It doesn't matter whether signal transduction through neurons is
energy efficient by the standards of computer scientists, or any other
standard. It is adaptive only because it is fast. A voltage change
can be propagated down a myelinated axon at 100 m/s. That's way
faster than any other type of signaling, like hormones, which is
dependent on diffusion rates or blood flow. That doesn't, however,
mean that it's the most efficient design that we could come up with if
we could manipulate the elements of neurochemistry ourselves.

Martin



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:56 MDT