How does the brain learn?

Summary: A new, open-source model of synaptic plasticity in the neocortex allows us to understand what learning is like in the brain.

A source: University of Montreal

Everyone knows that the human brain is very complex, but how does it learn? Yes, the answer may be much simpler than the general belief.

An international research team from the Université de Montréal has made great strides in accurately simulating synaptic changes in the neocortex, allowing them to better understand the brain.

An open-source study was published on June 1 Nature Communications.

“It opens up a world of new directions for scientific research on how we learn,” said Elif Mueller, assistant professor of IVADO research at UdeM and chairman of CIFAR AI in Canada, who led the study at the Blue Brain project at École University. polytechnic federation of Lausanne (EPFL), Switzerland.

Mueller moved to Montreal in 2019 and is conducting research on the architecture of the Biological Learning Laboratory, which he co-founded at the CHU Sainte-Justine Research Center with UdeM and Mila, Quebec Institute of Artificial Intelligence.

“Neurons are like trees, and synapses are the leaves of their branches,” said Mueller, the study’s lead author.

“The methods of the previous plastic model ignored the structure of this tree, but now we have computational tools to test the idea that synaptic interactions in branches play a key role in in vivo learning,” he said.

“It has important implications for understanding the mechanisms of neurodevelopmental disorders such as autism and schizophrenia, as well as for developing powerful new AI techniques inspired by neuroscience.”

Mueller, in collaboration with the EPFL’s Blue Brain Project, the Université de Paris, the Jewish University of Jerusalem, the Instituto Cajal (Spain) and a team of scientists from Harvard Medical School, developed a synaptic plastic model in a neocortex based on limited data. postsynaptic calcium dynamics.

How it works? It’s complicated, but in the end it’s simpler than you think.

The brain is made up of billions of neurons that communicate with each other to form trillions of synapses. These points of contact between neurons are complex molecular machines that are constantly changing as a result of external stimuli and internal dynamics, a process commonly referred to as synaptic plasticity.

In the neocortex, the main area involved in the study of high-level cognitive functions in mammals, pyramidal cells (PCs) make up 80-90 percent of neurons and are known to play a key role in learning. Despite their importance, the long-term dynamics of their synaptic changes have been experimentally characterized and varied among several types of PCs.

As a result, they had a limited understanding of the complex neural circuits they formed, especially the stereotypical cortical layers that defined the interactions of different regions of the neocortex.

The innovation of Mueller and his colleagues was to use a computational model to create a broader view of the dynamics of synaptic plasticity that governs learning in these neocortical schemes.

Comparing their results with the available experimental data, they showed in their study that their synaptic plastic model could capture different plastic dynamics of different PCs that formed the neocortical chip. And they did this using only a set of parameters of a single model, the plastic rules of this neocortex can be divided into pyramidal cell types and thus predictable.

Generalization test of the plastic model on the L2 / 3-PC connection type on the L4-PC. 3-D rendering of the L4-PC representation connected to the Silico model L2 / 3-PC. Inset shows an enlarged view of the mediator synapses (yellow balls). b Evolution of EPSP amplitude simulated during a conventional plastic induction protocol (upper left; one pair out of 100 is shown). Average EPSP amplitudes (top right) are shown before (base; blue) and after (long-term; orange) induction protocol. c Comparison of Silico and in vitro EPSP ratios for positive and negative times and with presynaptic NMDAR blocker MK801. Experimental data and simulations without MK801 (in vitro) on the right panel and MK801 on the left panel with γd = 0 (silico). The two-way t-test of Welch’s unequal variances was ns for each protocol (positive stimulus time from n-value: 0.268, 0.209 MK801, 0.959 MK801; n = 100). Experimental data (in vitro) Rodriguez-Moreno and Polsen42. Population data were reported as an average ± SEM. Credit: The Researchers

Most of these plastic experiments were performed in the in vitro brainstem of rodents, where the dynamics of calcium that control synaptic conduction and plasticity were significantly altered compared to in vivo intact brain learning. Importantly, the study assumes qualitatively different plastic dynamics from reference experiments conducted in vitro.

Mueller and his team believe that future experiments will have a profound effect on our understanding of plasticity and learning in the brain.

“The interesting point of this study is another confirmation for these scientists that gaps in experimental knowledge can be filled using modeling techniques in brain research,” said Henry Markram, EPFL neurologist and founder and director of the Blue Brain project.

“In addition, the model is open source, available on the Zenodo platform,” he added.

“Here we have shared hundreds of different types of plastic pyramidal cell connections. This is not only the most widely validated plastic model to date, but also the most comprehensive hypothesis of the differences between the plasticity and the intact brain observed in the Petri dish.

See also

This shows the person embracing the pillow and how the pillow is made

“This leap was made possible by our joint scientific approach. In addition, the public can take it further and develop their own versions by modifying or adding it – this will accelerate open science and progress. ”

News about synaptic plastic research

Author: Press service
A source: University of Montreal
The connection: Press Service – University of Montreal
Photo: The picture was given to the researchers

Original study: Open access.
Giuseppe Chindemi Nature Communications


Abstract

Calcium-based plastic model for predicting long-term potency and depression in the neocortex

Pyramidal cells (PCs) form the layered structure of the neocortex, and the plasticity of their synapses forms the basis of learning in the brain.

However, such long-term synaptic changes have been experimentally characterized between several types of PCs, making them a significant barrier to the study of neocortical learning mechanisms.

Here, we introduce a synaptic plastic model based on data-limited postsynaptic calcium dynamics and show that a set of parameters is sufficient to combine the available experimental results on long-term potency (LTP) and long-term depression in the neocortical chip model. (LTD) PC Connections.

In particular, we can explain the different plastic results for different types of PCs with specific synaptic physiology of the cell, cell morphology and innervation patterns, without requiring type-specific plasticity.

Summarizing the model in terms of extracellular calcium concentration in vivo, we assume a qualitatively different plastic dynamics from those observed in vitro.

This work provides the first comprehensive zero model for LTP / LTD between in vivo neocortical PC types, and a clear basis for further development of cortical synaptic plastic models.

Leave a Comment

Your email address will not be published.