Professor HyungGoo Kim's Team from the Global Biomedical Engineering Department Uncovers Principles of Instinctive Behaviors Using Artificial Intelligence

• Revealing the Mechanisms of "Hunger" and "Appetite" in Hypothalamic Neurons with AI-Based Models

Researchers led by Professor HyungGoo Kim from Sungkyunkwan University’s Global Biomedical Engineering Department (co-first author: master’s student Jong Won Yun) and Professor Hyung Jin Choi from Seoul National University (co-first authors: Ph.D. student Kyu Sik Kim, Dr. Young Hee Lee, and Ph.D. candidate Yu-Been Kim) have presented a novel method for understanding instinctive psychological states in humans through the application of artificial intelligence (AI).

Despite advancements in neuroscience enabling the observation of diverse animal behaviors, the connection between neural signals and instinctive psychological states remains insufficiently understood. Although prior studies have associated specific hypothalamic neurons with instinctive behaviors, the precise roles and mechanisms of these neurons were unclear. This collaborative research provides the first quantitative analysis of hypothalamic neural functions using AI, thereby clarifying the relationship between instinctive psychological states and behaviors.

The team combined a novel homeostatic theory with AI-based neural models to reveal that Agouti-related peptide (AgRP) neurons in the hypothalamus represent "hunger," while leptin receptor (LH LepR) neurons represent "appetite." Experimental observations of hypothalamic neural activity patterns were meticulously analyzed, experimentally demonstrating how hunger and appetite are encoded through the activity patterns of specific neural populations.

Professor Kim utilized a computational modeling approach, initially developed to distinguish the roles of dopamine, to devise a methodology for differentiating neural activations. This method successfully expressed Professor Choi’s novel homeostatic theory in mathematical terms.

Fusing Traditional Homeostatic Theories with Cutting-Edge Neuroscientific Discoveries

Professor Kim stated, “This study is the first to quantitatively analyze the activities of complex neural circuits by integrating AI with neuroscience. It marks a pivotal step toward numerically understanding instinctive behaviors in living organisms. Particularly, our elucidation of how hypothalamic neural activity regulates basic instincts like hunger and appetite holds significant implications.”

Professor Choi added, “Neural responses observed in the hypothalamus during studies on feeding behavior were difficult to explain using existing theories. AgRP neurons were activated to promote feeding but decreased their activity upon food presentation. Conversely, LH LepR neurons not only promoted feeding upon activation but also increased activity when food was provided. To understand these paradoxical findings, we employed AI models, which led to the establishment of a new homeostatic theory.”

Modeling Process Utilizing Artificial Intelligence

The study provides crucial insights into how the brain regulates survival-essential behaviors such as feeding. It is anticipated to pave the way for new strategies in treating eating disorders, obesity, and appetite-related conditions. By integrating AI into neuroscience, the research opens possibilities for quantifying and understanding human instinctive behaviors, offering a versatile approach to investigating other instinctive behaviors and psychological states.

Key Experimental Results

The study includes computer modeling to demonstrate how neural activity correlates with specific psychological elements. (Gray line: neural signals; red line: optimal hunger model; green line: optimal appetite model; blue and orange lines: control models).

The joint research by Professor Hyunggoo Kim and Professor Hyung Jin Choi was published in Science Advances on November 6.

• Paper Title: A Normative Framework Dissociates Need and Motivation in Hypothalamic Neurons
• Journal: Science Advances
• DOI: https://doi.org/10.1126/sciadv.ado1820