I recently finished up a two-year stint at a computational neuroscience research lab, which is part of a centre for theoretical neuroscience at a university. During my time there I developed a model of part of the brain.
What does that all mean? Let’s find out.
Note: This post is just meant to be a brief introduction; I’ll be simplifying things with the aim of getting the main ideas across.
Most scientists spend their time developing theories and/or performing experiments. The two don’t exist in isolation: before you start an experiment you usually have some idea (theory) about what’s going to happen, and the results obtained from experiments are needed to validate or disprove theories.
People in the field of theoretical neuroscience come up with ideas about how the brain works: how it processes visual data, how it coordinates the movement of muscles, etc.
The particular theory I worked on is related to how brains (mammalian brains) learn in fearful situations. For example, if you give a rat an electrical shock in a particular environment, the next time he goes back to that environment he’ll freeze up and his heart will start beating fast: similar to what happens when we are afraid. I made informed guesses as to what might be going on in the brain that allows the rat to learn that he should be afraid in that particular environment.
To test the theory, I looked at the results of experiments that other people had performed. Does my theory explain all the different behaviours we see from rats when we put them in situations that involve things they’re afraid of? Does my theory include the same brain regions that ‘light up’ when rats are exposed to things they’re afraid of? The answer is usually ‘kinda’, so I was always tweaking my theory based on some new experimental results in order to make it better. That’s generally how theoretical neuroscience works.
The brain is a computer. No, it doesn’t work like our computers at home, but it performs a similar task. It’s an information processing machine: it takes inputs and generates outputs. Figuring out how our brains do this is the main focus of the field of computational neuroscience.
In order to process information, a computer needs a way to represent information. Our computers at home use transistors and electrical activity to do this. When you type ‘1+2’ into a computer, ‘1’, ‘+’, and ‘2’ are all represented as electrical activity in the computer. That information is processed in the computer (electrical activities are combined) to generate electrical activity that results in an output of ‘3’ to the screen.
Our brains do something similar, but they use biological cells (neurons) and electrical activity to represent information. If you happen to see a scary man holding a knife, ‘scary man’, ‘holding’, and ‘knife’ are all represented as electrical activity in your brain. That information is processed in the brain (electrical activities are combined) to generate electrical activity that results in your muscles moving you very quickly in the opposite direction of the man.
The challenge is to use math to describe the information processing going on. With computers, we say that when a transistor has some electrical activity its value is 1, and when it has no electrical activity its value is 0. The computer adds together the 1s and 0s to generate a new output. But what about the brain? Can we classify electrical activity on neurons as 1s and 0s and say that the brain is adding them up? Turns out it’s more complicated than that. But the general idea is the same: come up with a numerical representation for the information and describe what the brain is doing as a mathematical operation on that information. That’s computational neuroscience.
If you can describe something mathematically, you can write a computer program to simulate it. People use computer programs to simulate all sorts of things. A flight simulator (a model of an airplane flying) is a good example. It represents wind, elevation, speed, and rudder positions with numbers and performs mathematical operations on those numbers to show how the airplane would behave in that situation.
Similarly, a brain simulator (a model of a brain or part of a brain) represents light, sound, smell, taste, and touch with numbers and performs mathematical operations on those numbers to show how the animal would behave in that situation.
Why do we want to make a model of the brain? Here are a couple good reasons.
One is so that we can have software that does the things that brains do. For example, brains can rapidly process changing visual scenes and generate complex motor plans that allow you to, say, drive a car on a busy street. If we could capture that processing in software, then we could let that computer program drive our cars for us.
Another reason to model brains is so that we can do things like simulate brain diseases in the model and then simulate the effects of drugs on those diseases. This means that we could test out treatments for diseases on a computer program before we test them out on animals or people.
No, we don’t have detailed models of how the brain is able to drive a car or how certain diseases and drugs affect the brain… yet. But, I feel that this pursuit of replicating in software the way that the brain processes information is going to lead to some really exciting technologies and medical applications in the future.