The modularity and robustness of the Cell brain

Responsible Editor | How do people and animals maintain a high degree of robustness (robustness) to survive in a complex and changeable environment full of noise? What is the brain neural mechanism behind it? On July 1, 2021, Dr.
Guang Chen, Dr.
Nuo Li () of Baylor College of Medicine and PhD students Byungwoo Kang and Dr.
Shaul Druckmann of Stanford University published an article on Cell titled Modularity and robustness of frontal cortical networks, Uncovered the neural mechanism of the brain to maintain robustness (robustness, robustness)
.

The cognitive and motor functions of the brain weave a variety of behaviors of humans and animals
.

One of the most important functions is short-term memory, which connects the past, present, and future moments to form a coherent behavior
.

For example, if someone tells you your name or phone number, you can immediately remember it for a while and store it in your mobile phone address book
.

Past studies have found that the continuous electrical nerve activity in the brain is closely related to short-term memory function, and is considered to be the basis of neural activity supporting short-term memory
.

This kind of continuous electrical activity exists in many different brain regions.
It is generally believed that the cooperation of multiple brain regions maintains continuous electrical activity and short-term memory [1, 2]
.

At the same time, people also believe that multi-brain cooperation can better combat noise disturbances in the internal and external environments, because redundant information in other brain areas can compensate or remedy the information in the disturbed brain area
.

But another prerequisite for this cooperative compensation is that different brain regions need to maintain a high degree of independence (modularity) when they are locally disturbed to prevent the destructive information of the disturbed brain region from spreading to other undisturbed areas.
The brain area thus loses the opportunity to be compensated
.

However, different brain areas in the brain are highly connected structurally, thus giving opportunities for destructive information to spread
.

So, how can neural networks in different brain regions not only communicate and cooperate with each other, but also prevent the spread of disturbances so as to maintain a robust short-term memory function? The laboratory’s past research has found that mice can resist unilateral but not bilateral perturbations in the prefrontal cortex of the brain when doing short-term memory behavior tasks to maintain normal behavior [3]
.

This discovery establishes a model animal system for studying the robustness of cognitive functions, and suggests that other non-disturbed brain areas, such as the undisturbed contralateral prefrontal cortex, can compensate for the disturbed brain areas to maintain robustness
.

So, when a unilateral disturbance occurs, does the contralateral brain area really maintain modularity and can remedy the entire system to achieve robust behavior after the disturbance? How do the two brain areas interact? To answer this question, the authors in this study performed recordings of the electrical activity of large-scale neuronal populations and optogenetic disturbances in the prefrontal cortex on both sides of the brain when the mice were performing short-term memory behavioral tasks
.

The experiment found that the prefrontal cortex on both sides simultaneously represented two similar short-term memory information, and the neural activity that represented this information showed a high degree of consistency (Figure 1)
.

Further experiments and video analysis of mouse movement suggest that the coordination of activities on both sides is likely to be formed by the interaction of the neural networks on both sides, rather than the common input of other brain regions
.

Figure 1.
Coordinated neural activity of the two prefrontal cortex in short-term memory.
To verify this, the authors directly measured the interaction between the two brain regions
.

They used optogenetic methods to directly inhibit the neural activity of the unilateral prefrontal cortex in the early stages of the memory stage, and at the same time observed the degree of influence of the disturbance on the contralateral brain area
.

It was found that some mice showed a high degree of modularity, that is, inhibiting the neural activity of the left side had almost no effect on the right side, and inhibiting the neural activity of the right side had almost no effect on the left side, presenting two modules (Figure 2)
.

Other mice lack modularity.
The left side strongly dominates the activities on the right side while the right side does not affect the left side, presenting an asymmetric single module
.

In general, different mice have different degrees of modularity, showing approximately continuous changes
.

Further analysis revealed that the strongly asymmetrical single module showed a high degree of consistency in the activities on both sides, while the relatively independent dual-module system showed a lower consistency
.

It shows that the coordination of activities on both sides is indeed related to the interaction of the cortex on both sides
.

Further experiments found that this differentiated and left-sided dominance interaction between the two sides of the brain in different mice is not completely innate.
Changing the structure of behavioral tasks can reverse the left-side dominance to the right-side dominance
.

Figure 2.
Different types of interactions between the two sides of the prefrontal cortex.
Given that different mice have different degrees of modularity, do mice with high modularity have better robustness? Further analysis showed that highly modular mice can restore normal short-term memory no matter which side of the brain is disturbed.
For mice with high asymmetric and low modularity, disturbing the dominant side strongly destroys their short-term memory.
Memory (Figure 3)
.

Figure 3.
Modularization enhances the robustness of short-term memory.
The above results indicate that the low-interaction modularity of the brain regions prevents the diffusion of damaged information (modularity) is necessary for robustness, but at the same time, it does not disturb the normal memory information in the brain area.
It is necessary to remedy (error-correction) the information that disturbs the damaged brain area through the strong interaction of the brain area
.

How can these two similarly contradictory requirements be realized in the same neural circuit connection at the same time? A simple solution is that one brain area can switch the intensity of the opposite side to it according to the state of its neural activity encoding memory information
.

When it encodes a specific memory with high selectivity, it reduces the influence of the opposite side on it, and vice versa, it allows the opposite side to influence it
.

Based on this hypothesis, the authors modeled and analyzed the undisturbed neural activity on both sides to infer the strength of the interaction between the two sides, and found that the hypothesis was indeed valid
.

Later, a new experiment was designed to further prove this hypothesis (Figure 4)
.

This state-dependent gating mechanism, which exerts force on the two brain regions on the same neural circuit, can explain the modularity and error correction of the circuit well at the same time
.

More importantly, this state depends on the degree of gating to predict the performance of mice in normal behavioral tasks
.

Figure 4.
The state of inter-brain interaction depends on gating adjustment.
Finally, the authors explored what kind of network architecture (brain-interval action mode) can be formed by training recurrent neural network (RNN) to perform the same behavioral task The solution of a robust network
.

And under what conditions, the phenomena and principles observed in the experiment can be repeated in the artificially trained neural network
.

Studies have shown that for a neural network to have a robust performance in this short-term memory task, three basic conditions need to be met, namely, the modularized initial connection between the cortical networks on both sides, modularized training, and the nonlinear characteristics of neurons
.

In this trained neural network, the state-dependent gating phenomenon between the networks on both sides also emerges at the same time
.

And similar to experimental research, changing the relative intensity of the network sensory input on both sides of the training can change the symmetry structure of the artificial neural network
.

Figure 5.
Working model.
In general, the study found that the prefrontal cortex on both sides of the brain encodes short-term memory information in parallel, and it can be realized by different ways of inter-brain interaction
.

But the modular mode of action enhances the robustness of the brain and promotes behavioral performance
.

The same neural connection in the brain area dynamically adjusts the degree of interaction through a state-dependent gating mechanism to achieve modularity and error correction (coordination) at the same time
.

The modular architecture also exists in the robust artificial neural network for learning short-term memory tasks (Figure 5)
.

The modular architecture that is neither too unitary nor too decentralized to enhance the robustness of the system is not only found in the brain, but also exists in other network systems
.

For example, when one part of the power network is destroyed, it can cut off its influence on other parts, and in turn, other normal parts can help destroy the power supply of the area
.

Another example is the social population network system.
When the virus is raging today, a simple and effective way for people to resist the virus is to cut off the source of infection’s impact on the uninfected people.
At the same time, the uninfected people actively treat the infected people to reduce humanity.
Loss
.

The modular architecture found in the study to enhance the robustness of the brain may be used as one of the general principles to inspire future research on artificial intelligence and the development of neuromodulation technology for the treatment of neuropsychiatric diseases
.

Original link https://doi.
org/10.
1016/j.
cell.
2021.
05.
026 Platemaker: Eleven References 1.
Christophel, TB, Klink, PC, Spitzer, B.
, Roelfsema, PR, and Haynes, JD ( 2017).
The Distributed Nature of Working Memory.
Trends Cogn Sci 21, 111-124.
2.
Svoboda, K.
, and Li, N.
(2018).
Neural mechanisms of movement planning: motor cortex and beyond.
Current opinion in neurobiology 49, 33-41.
3.
Li, N.
, Daie, K.
, Svoboda, K.
, and Druckmann, S.
(2016).
Robust neuronal dynamics in premotor cortex during motor planning.
Nature 532, 459–464.
Instructions for reprinting [not original Article] The copyright of this article belongs to the author of the article.
Personal forwarding and sharing are welcome.
Reprinting is prohibited without permission.
The author has all legal rights and offenders must be investigated
.