Revolutionizing ALS Research: AI Models Predict Neural Network Degeneration
A groundbreaking study from the University of St Andrews, the University of Copenhagen, and Drexel University introduces AI computational models that can predict the degeneration of neural networks in Amyotrophic Lateral Sclerosis (ALS).
Published in Neurobiology of Disease, this research opens up new possibilities for using computational modeling as a complementary approach to traditional animal and in vitro methods. Motor neuron disease (MND), a term encompassing various illnesses affecting motor neurons in the brain and spinal cord, is primarily known by its most common subtype, ALS, which affects approximately 2 out of 100,000 individuals annually worldwide, and around 200 people in Scotland each year.
ALS often begins with spinal onset, where motor neurons and specific neural circuits in the spinal cord are initially affected, leading to early symptoms like muscle weakness, stiffness, and cramps. Traditionally, ALS research has relied on animal models, such as genetically modified mice, to study disease progression. However, these models have limitations, as researchers can only focus on specific time points due to time and cost constraints.
Here's where AI computational models shine. They can predict the disease's progression between these time points, offering a more comprehensive understanding. Moreover, these models can be easily modified to understand the impact of specific changes, whereas animal models are influenced by numerous factors. Importantly, they can also predict how neural circuits might respond to treatment, providing valuable insights for future preclinical studies in mice.
The study's researchers utilized biologically plausible neural networks, distinct from those used in everyday tasks like facial recognition or ChatGPT. These networks communicate using spike signals, mirroring the nerve cells in our nervous system. The networks are structured based on known spinal cord cells and their connections, allowing researchers to develop models grounded in biological knowledge.
These models, created by the School of Psychology and Neuroscience, are mathematical equations calculating neuron excitability. When a neuron receives a spike (an electrical impulse), its excitability changes, and if it reaches a threshold, it spikes, passing information to the next neuron. Neurons are grouped into populations and connected based on biological data to construct the network.
Co-author Beck Strohmer, a postdoctoral researcher from the University of Copenhagen, explains that during ALS, neurons die, and communication between populations breaks down. They model this by removing neurons from affected populations and reducing connections. This enables them to predict disease progression and test treatment strategies by saving or strengthening communication.
Co-author Dr. Ilary Alodi, a Reader in St Andrews School of Psychology and Neuroscience, emphasizes the need to test hypotheses generated by models on animal models due to the complexity of biological systems. In this study, they predicted a treatment strategy's effectiveness in saving a specific neuron population, which was confirmed in treated mice.
These findings highlight the potential of AI models to guide experimental research while exercising caution with predictions. This refinement in animal experimentation allows researchers to pinpoint changes in animal models more effectively.
Dr. Alodi adds that they are now applying these models to specific brain areas to understand neuronal communication changes during dementia, opening up exciting new research avenues.
