Molecular computation in neurons: a modeling perspective

- Most neuronal computation is chemical, not electrical.
- Chemical networks decode stimulus patterns to trigger distinct cellular outcomes.
- Plastic changes can be stored long past the turnover time of individual molecules.
- Cascades of molecular switches respond quickly and store information for long times.

Neurons perform far more computations than the conventional framework of summation and propagation of electrical signals from dendrite to soma to axon. There is an enormous and largely hidden layer of molecular computation, and many aspects of neuronal plasticity have been modeled in chemical terms. Memorable events impinge on a neuron as special input patterns, and the neuron has to decide if it should 'remember' this event. This pattern-decoding decision is mediated by kinase cascades and signaling networks over millisecond to hour-long timescales. The process of cellular memory itself is rooted in molecular changes that give rise to life-long, stable physiological changes. Modeling studies show how cascades of synaptic molecular switches can achieve this, despite stochasticity and molecular turnover. Such biochemically detailed models form a valuable conceptual framework to assimilate the complexities of chemical signaling in neuronal computation.