Included here is a summary of the following article:
Bhalla, Upinder S. "Molecular computation in neurons: a modeling perspective." Current Opinion in Neurobiology 25 (2014): 31-37.
Introduction
- Chemical signaling probably accounts for a hundred times as much computation as electrical signaling.
- What is involved in these computations? There are over 1400 distinct proteins in the post synaptic region.
- Chemical computation includes:
- presynaptic signaling
- bistability from chemical feedback loops
- bistability from receptor trafficking
- 'off' state with few receptors
- synaptic tagging and local activity spread
- spine structural change in plasticity
- input patterns control plasticity
- activity controls gene expression
Review
Pattern selectivity
- Calcium influx can trigger both increased or decreased synaptic efficacy, and there may be local as well as global differences throughout the neuron
- Mitogen-activated protein kinase (MAPK) cascade shows preference for bursts that are spaced by 5-10 min
- Kinase cascades have the ability to respond to a range of time-scales from seconds to minutes
- After all, the spike-timing dependent plasticity (STDP) is a direct result of channel biophysics
Spatial models and spatial pattern selectivity
- Synaptic tagging - strong activity at one synapse enables the potentiation of a nearby synapse receiving much weaker input
- Reaction-diffusion models demonstrate how long term potentiation and long term depression spreads to neighboring spines
- The spatial spread of calmodulin (CaM) plays a role in decoding synaptic input patterns
Signal Propagation
- Precision in spatial signaling can be achieved through chemistry via diffusion but perhaps through other complex means
- Chemical signaling might support long range signal propagation such as through active transport via cellular motors
- Many chemical computations can occur in parallel in each neuron
Long-term storage: biochemical bistability
- Three processes act against the stable long-term storage abilities of synapses: turnover, diffusion, and stochasticity.
- It has been showed that bistability, which is often a demonstration of how a positive feedback loop may strengthen memories, has a very narrow range in synaptic/dendritic protein synthesis.
- There are many biochemically plausible candidates for memory storage at the synapse.
Plasticity and molecular traffic
- How does a synapse retain its state given diffusion processes that occur?
- One idea is molecular crowding that can preserve receptors in the synapse for periods of hours
- It is the molecular movement of receptors as opposed to their signaling identity that is needed for LTP, for example.
Losing state: stochasticity and bistability
- Given how small a synapse is, and that there are only one to two hundred calcium receptors, there are strong fluctuations in chemical activity. Such "noise" affects state transitions.
- These stochastic fluctuations may even be sufficient to cause state flips in biochemical bistables, erasing any memories they store. Many bistable models fail the stochasticity test.
Continuous synaptic strengths
- Most studies report continuous plasticity (as opposed to bistable states) without explicitly addressing whether this is at the single synapse level.
- Continuous states could be accomplished by having many stable states, which are indeed possible, but only through using many bistable states to accomplish a multistate system.
Conclusion
- Most computation at the neuron level is chemical, such as the roles of pattern decoding and memory storage
- It may be advantageous to use electrical computation when it is a higher level that encompasses much low level chemical computation