Building a computer system that can replicate the human brain’s ability to learn new tasks has been scientists’ dream for decades. MIT researchers are now one giant step closer to realizing this dream by designing a computer chip, which mimics how the brain’s neurons adapt in response to new information. This process, known as plasticity, is believed to be the basis of many brain functions, such as memory and learning. The findings will be described by senior author Chi-Sang Poon at the National Academy of Sciences this week.
The silicon chip, which has approximately 400 transistors, is able to simulate the activity of a single brain synapse. A synapse is a link between two neurons, cells that are specialized to pass signals to individual target cells, and a synapse is the means by which they do so.
Chi-Sang Poon, principal research scientist in the Harvard-MIT Division of Health Sciences and Technology, says they expect this chip to assist neuroscientists in learning more about how the brain functions. The chip could also be utilized in neural prosthetic devices like artificial retinas.
The brain contains about 100 billion neurons, each forming synapses with many other neurons. A synapse is the cleft (gap junction) between two neurons, one is called the presynaptic neuron and the other one the postsynaptic neuron. The presynaptic neuron releases neurotransmitters, i.e. glutamate and GABA, which bind to the postsynaptic cell membrane’s receptors, activating ion channels that are capable of passing electrical current.
The cell’s electrical potential is changed, by opening and closing of these ion channels, which causes voltage changes in the presynaptic cell to induce voltage changes in the postsynaptic cell. This electrical impulse is called an action potential.
The entire synaptic activity depends on these ion channels that control the flow of charged atoms, i.e. sodium, potassium and calcium. These ion channels are fundamental for two processes, such as the long-term potentiation (LTP), which strengthens synapses and the long-term depression (LTD), which weakens synapses.
The MIT researchers designed the computer chip in such a way, that the transistors could mimic the activity of different ion channels. In comparison to most chips that operate in a binary, on/off mode, the researchers designed the new brain chip so that current flows through the transistors in analog and not digital fashion, with a gradient of electrical potential driving the current to flow through the transistors similar to ions flowing through ion channels in a cell.
Poon comments:
“We can tweak the parameters of the circuit to match specific ion channels. We now have a way to capture each and every ionic process that’s going on in a neuron.”
Before, researchers used to build circuits that could only simulate the firing of an action potential without all the circumstances that produce the potentials. Poon adds:
“If you really want to mimic brain function realistically, you have to do more than just spiking. You have to capture the intracellular processes that are ion channel-based.”
The MIT researchers anticipate their chip to be used for building systems to model specific neural functions, like the visual processing system, which could be significantly faster compared to digital computers. To simulate a simple brain circuit can take hours or days, even on high-capacity computer systems, whilst simulation with the analog chip system is even faster than the biological system.
According to Poon, building chips that can interface with biological systems could be another potential application, which could benefit a potential communication between the brain and neural prosthetic devices, like artificial retinas. In the far future these chips could also serve as building blocks for artificial intelligence devices.
The chip has already been used by the MIT team in order to propose a resolution to a longstanding debate over how LTD occurs.
According to one hypothesis, LTD and LTP depend on the frequency of action potentials stimulated in the postsynaptic cell. A more recent theory however suggests that the timing of the action potentials’ arrival at the synapse is the key on which long-term potentiation and long-term depression depend upon, as both require the involvement of ion channels known as NMDA receptors, which detect postsynaptic activation.
Another recent theory suggests that both models could be unified if a second type of receptor would be involved in detecting that activity. One potential candidate for the second receptor is the endo-cannabinoid receptor.
Endo-cannabinoids, comparable to the structure of marijuana, are produced in the brain and play a role in various functions, such as pain sensation, appetite and memory. According to some scientists’ theories, if endo-cannabinoids, which are produced in the postsynaptic cell, are released into the synapse, they activate presynaptic endo-cannabinoid receptors, and LTD will occur, if NMDA receptors are active at the same time.
The researchers managed to accurately simulate both LTD and LTP by including transistors in their chip, which modeled endo-cannabinoid receptors. Poon declared that despite the support of previous experiments, until now, “nobody had put all this together and demonstrated computationally that indeed this works, and this is how it works.”
The study was led by Chi-Sang Poon, principal research scientist in the Harvard-MIT Division of Health Sciences and Technology, with Guy Rachmuth, a former postdoc in Poon’s lab, as a lead author. Other authors include Mark Bear, the Picower Professor of Neuroscience at MIT, and Harel Shouval of the University of Texas Medical School.
Written by Petra Rattue