The brain is considered the most complex organ in the human body with approximately 86 billion neurons. Its ability to interpret senses, coordinate body movement and control behavior make it an incredible living and learning organ.
However, according to the World Health Organization, 317 million children in 2019 were affected by health conditions attributed to neurodevelopmental disorders, including autism. Computer Science and Engineering Assistant Professor Christian O’Reilly recently received a more than $600,000 National Science Foundation CAREER Award to develop a framework to support multiscale model-driven analysis of the brain and its disorders.
“I’ve worked on addressing problems we have with understanding the brain,” O’Reilly says. “By studying different regions, you can see how their activity correlates with some processes and how they are affected due to a condition such as a stroke.”
O’Reilly intends to establish a modeling framework for addressing issues related to unusual neural activity in autism. This refers to changes in brain regions that may lead to symptoms such as difficulties in social cognition and restrictive and repetitive behaviors.
The brain processes spatial information across three main scales: small, medium and large. This refers to different levels of organization in the brain, from the neuron at the smaller scale to the entire brain at the larger scale.
“Without a common framework to integrate all this data, it's hard to have a complete picture,” O’Reilly says. “Modeling offers the opportunity to integrate data across scales.”
O’Reilly plans to focus on the cortex, which is the largest and most recently evolved part of the brain. It is crucial to higher cognitive functions, including memory, reasoning and sensory perception. In first approximation, the cortex can be considered as being built up of a large number of cortical columns which consist in functionally segregated networks of neurons and are thought to form one of the basic functional unit of the cerebral cortex.
“The project will simulate what we call a canonical microcircuit of the cortical column, a blueprint for this essential building block of the cortex,” O’Reilly says. “We’ll simulate how a population of neurons come together within these cortical columns.”
O’Reilly’s project will develop a novel model-driven approach, focusing on three aims to build the framework and apply it to the specific use case: autism research.
“In the context of autism, there’s a research direction that studies an imbalance between two types of neurons in the brain: excitatory neurons, which tends to cause other neurons to generate neural spikes and inhibitory, which serve as the brakes,” O’Reilly says.
According to O’Reilly, when studying human brains with a non-invasive approach, many important variables cannot be easily measured, as in the case for the balance between excitatory and inhibitory neurons. But having a model which defines how variables may cause observable brain activity, experimental data can be used to infer from these non-observable variables.
Often, we don’t always understand why people have autism, and in most cases, we can’t pinpoint a specific mutation.
- Christian O'Reilly
The first aim will produce brain simulations by designing what O’Reilly refers to as forward model, with the goal of better understanding how the brain generates observable signals, such as from an EEG.
“It’s determining how different activations are generated in the brain. It’s similar to the concept of digital twins, where we reproduce a facsimile of the real system in a computer, which in this case is the brain,” O’Reilly says. “It involves a cross-scale integration since we’re trying to capture information from the molecular exchange to the whole brain activity.”
The second attempts the inverse of the simulation by trying to solve an inference problem. From observable recordings, it attempts to determine the value from hidden variables.
“You go from the final signal and figure out the values for the different parameters in the model required to reproduce these signals. Inverse modeling will allow us to do inference on these parameters,” O’Reilly says. “For example, if I have an EEG recording, I can use this inference process to determine the balance between excitatory/inhibitory neurons and their activity.”
The third aim will apply the cross-scale integration in the simulation and the parameter inference method to the specific use case of autism. O’Reilly will work with Caitlin Hudac, a professor in the University of South Carolina Department of Psychology, who has performed research on people with autism with different genetic mutations. The results will allow O’Reilly to estimate the values of the parameters from EEG signals and investigate whether the imbalance in excitatory/inhibitory ratio in autism is associated with a breakdown in long-range effective connectivity.
“Often, we don’t always understand why people have autism, and in most cases, we can’t pinpoint a specific mutation. But she has a special data set from people with four different rare genetic mutations,” O’Reilly says. “These four groups of people with autism have overlapping symptoms, but the causes are different. The question is, ‘How are these genetic causes affecting the neural mechanisms, and can we infer these causes non-invasively with the approach that I’m developing?’”
O’Reilly is also excited about the educational component of the CAREER Award. He plans to develop a set of educational videos on computational neuroscience, an interdisciplinary field that utilizes mathematical models, theoretical analysis and computer simulations to understand the structure and functions of the nervous system and brain.
“It's a challenging field because students generally come from engineering, computer science, psychology or biology but rarely have both the math and programming skills as well as an understanding of how the brain and neurons work,” O’Reilly says. “This will make it easier to involve undergrads and grad students in our research.”