Tuesday, May 22, 2012

Neurons are like equations

The brain of the clock (I took this picture)
A computational model is a surrogate version of something usually made on a computer.  An example that most people are familiar with are the computational models used to predict the weather. If you know how low pressure and high pressure fronts interact, and you know where one is and how fast it is moving, you can program software to play the situation out in a simulation, predicting what will happen and how quickly. 

Computational neuroscience is more or less just like that and it can be used to investigate all levels of neuroscience. Here's a brief intro to three of the basic levels. There are other types of computational models in neuroscience, but these three make up most of them.

The Whole Brain
If you know how the thalamus, hippocampus, amygdala, and cortex all work together, you can simulate how inputs into one structure might influence the others. In this case each brain structure would basically be a 'black box' that received input and produced output based on known data.  To do this kind of simulation you wouldn't actually simulate the millions of neurons in each structure.

The Neural Network

On the next level down, you can make a computational model of a neural network inside a single brain structure. If you know the types of neurons in the amygdala and how they interact with each other, you can program those relationships in and test what might happen if one class of neurons fires too much or too little. You can test the effect removing one class of neurons has on the whole network and the output of that brain structure.  In this case you are simulating individual neurons, but you are probably not simulating the details of the neurons, such as their dendrites and their specific channel composition. In this kind of computational model, the neurons are the 'black box' which receive input and produce output based on pre-set equations.

The Cellular Scale
One level down from this is a computational model of an individual neuron. In this type of model, the neuron is simulated in detail, with its dendrites, soma, and sometimes the axon.  With this kind of model, you can test the effects of different dendrite shapes on the processing of the neuron. Usually the individual channels (such as calcium, potassium and sodium channels) in the neuron are programmed in and the electrical properties of the cells are calculated in detail. In this situation, the specific proteins and channels are the 'black boxes' computing ionic concentrations based on pre-set equations. A detailed tutorial on how to make a biophysically realistic model neuron can be found here.

a neuron can be simulated as a series of resistors and capacitors

Sidiropoulou et al., (2006) have an excellent review  of the neuroscience discoveries that have been made with this cellular level of computational modeling.

They start their paper highlighting the most interesting problem in cellular neuroscience.

"Understanding how the brain works remains one of the most exciting and intricate challenges of modern biology. Despite the wealth of information that has accumulated during the past years about the molecular and biophysical mechanisms that underlie neuronal activity, similar advances have yet to be made in understanding the rules that govern information processing and the relationship between the structure and function of a neuron." (Intro, Sidiropoulou et al., 2006) (red mine)
This paper directly argues against the idea that neurons are just 'on-off' switches, and illustrates the complex computational processes that occur in individual locations of the neuron. They cover computational studies analyzing the information processing that occurs in the dendrite, at the synapse, at the soma, and even in the axon. The details are to complicated to get into here, but the paper is free.

Finally, they end with a call to action for experimental and computational neuroscientists to work together to solve the really interesting problems in cellular neuroscience. 
"The following open questions could provide fertile ground for collaborations among molecular biologists, geneticists, physiologists, modellers and behaviourists for further explorations of the mysteries of the brain. Do specific behaviours require certain neuronal computational tasks? Which parts of the neural circuit or the neuron itself are responsible for these tasks? What are the underlying molecular mechanisms for the distinct operating modes of neuronal integration? Such holistic approaches should lend support to the growing idea reinforced by this review: that something smaller than the cell lies at the heart of neural computation." (Discussion, Sidiropoulou et al., 2006)
Just as computational models can predict weather patterns with some degree of accuracy, no model is perfect.  Similarly computational neuroscience is not going to lead to all the answers, but where it is particularly useful is in making very specific predictions about how certain aspects of a neuron or neural circuit might work. The insight gained from computational models can guide and focus experiments, making them more efficient. This saves time, money, energy, and animal lives.

© TheCellularScale

Sidiropoulou K, Pissadaki EK, & Poirazi P (2006). Inside the brain of a neuron. EMBO reports, 7 (9), 886-92 PMID: 16953202


  1. It is not just the complexities of neural cells that has to be investigated, but also the complex nature of the support cells. The whole system is responsible for brain activity, change and healing of damage. There is less know about the support cells than neurons. http://www.cell.com/neuron/fulltext/S0896-6273(08)00886-6

    William Fairholm

  2. Great point. Both the Glia and the inhibitory interneurons play a huge role in neural activity and brain function. Computational neuroscience is starting to look at those, and the 'whole network' level of research is leading to really interesting insights. But so so so much more work needs to be done.