X-Message-Number: 26333 From: Date: Tue, 14 Jun 2005 13:43:23 EDT Subject: Uploading technology (2.vi.1) Presynaptic model. Uploading technology (2.vi.1) Presynaptic model. The traditional view is that different effects on the presynaptic side, the axon terminal, encode the memory. I think this is not the case, at least for long term memory. My idea is that this one is encoded in the dendrite spines working as en route signal amplifiers. So, what is the presynaptic neurotransmitter modulation good for ? It has been demonstrated that synapses with two different time constants can display a rich set of behavior (*). This would correspond to two different currents at the membrane level. With N neurons, it is possible to encode at most N-1 different patterns. A robust system with large redundancies would have room for far less yet. The general idea is that pattern P1 regulated by the first current is progressively put out of phase in the network by the second current. So the system skip to pattern P2 and so on. If the system is Markovian, that is it rests for the next pattern only on the preceeding one, then there is a number of disconnected cycles: For example P1, P2, P4, P1, P2, P4, P1,... and P3, P5,P3,... If there is some dependency on more than one Pn then there could be: P1,P2,P3,P1,... and P3,P4,P5,P3. Here, P3 is included in two cycles, but what come next, P1 or P4 rests on what was the state before P3, either P2 or P5. Here, P3 is a bridge between two cycles, it can define an associative idea between them. I think a similar process can take place even with a single current. In most case neurons don't work with single pulse, they use large pulse trains and these can produce complex interactions from pulse to pulse in the same train. There is four positive interactions at the presynaptic level. If a pulse follow another after a short time span, it is amplified, that is, the propability Pro of producing a neurotransmitter release is augmented. There can be short time effects with large amplitude or long time one with smaller amplitude for each impulse. The 4 positive interactions are labeled first facilitation, second facilitation, augmentation and potentiation. The decay time is (**): First facilitation (F1): 50 ms, 80% more magnitude for the second impulse. Second facilitation (F2): 300 ms, 12% more. Augmentation (A): 7s, 1%. Potentiation (P): from 20 seconds to some minutes, 1% or less. There is too a contrairy effect, the depression with at least two components: The fast one (D1): 5s, from 0 to -15%. The slow one (D2): Up to some minutes, from 0 to 0.1%. The ratio between the effective membrane potential and the basic one at time t (in ms) is, for facilitation only: 0.8exp -(t/50) }+ 0.12 exp -(t/300) + 1 The factor 0.8 and 0.12 stand for the 80 % and 12 % in the above list and t/50, t/300 come from the decay time in milliseconds. It is simple to complete the formula with augmentation, potentiation and depression, fast and slow for complete decay behavior. Looking now at how the effect build up, two formulas seem good: (F1+F2+1)^3 (A+1) (P+1), and: (F1+F2+A+1)^3 (P+1) It is disturbing at the research level that two different models satisfy the experimental data and this is here a problem. For uploading purpose there is no difficulty, either one can be used. A pattern could be formed by a set of short pulses so that only first facilitation would be activated. With repetitive patterns yet, the second facilitation would enter into play and shift the system to another pattern. This would work as if there was two currents acting on individual pulses. The pattern cycling can then evolve along two dimensions: One resting on different currents and acting at the single pulse level and another working on pulse trains and using the facilitation 1 and 2, the augmentation, potentiation dans short-long depression. At least for uploading purposes, I suggest we can concentrate the F1,F2,A,P,D1,D2 on the presynaptic, axon side and the currents effect on the postsynaptic, dendrite side. (*) Kleinfeld D., Sompolinsky H. in Methods in Neuronal Modeling Koch C. and Segev I. ed. MIT Press. (**) MAGLEBY K. L., in Synaptic Function Edelman G.M., Gall E.W. and Cowan W.M. ed. John Willey and Sons. Yvan Bozzonetti. Content-Type: text/html; charset="US-ASCII" [ AUTOMATICALLY SKIPPING HTML ENCODING! ] Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=26333