Summary: Machine learning is helping researchers uncover the genetic influence on foraging behaviors in mice.
Source: University of Utah Health
Mice scurry while searching for food, yet their wanderings may be shaped by genes. Researchers at University of Utah Health applied machine learning to connect genetic variation with the incremental steps that compose instinctive and learned behaviors. Their findings were published in Cell Reports on August 13.
“Complex behaviors like searching for food appear spontaneous and unpredictable, but they are made of repeating sequences,” said Christopher Gregg, Ph.D., assistant professor of Neurobiology and Anatomy at the University of Utah Health and senior author of the study. “With machine learning, we can identify discrete behavioral sequences that occur more often than expected by chance, and these sequences have biological roots.”
The team is exploring the architecture of behavioral sequences to reveal how genes shape patterned actions.
“Our goal is to map how complex behavior is constructed and how genetic differences alter those patterns,” Gregg said.
The study supports a model in which complex behavior is assembled from a finite set of “behavioral modules” — reproducible building blocks whose order, timing and frequency are under genetic control. Different combinations and transitions among these modules create distinct behavioral patterns.
To test this idea, researchers observed 190 mice with variations in genetics and age as they moved from their home base into a specially designed arena and foraged for food. Foraging engages many neural systems simultaneously, including those that govern exploration, anxiety, reward, energy balance, attention, navigation and memory. The new analytical methods showed that genetic background and age differentially affect specific sequences of behavior.
“Most animals organize activity around a home range, and we found reproducible behavioral sequences tied to that structure,” Gregg said. “Identifying these sequences helps us describe behavior across time with much finer detail.”
Investigators decomposed round trips from home to the food source and back into more than 5,600 discrete mouse actions. Each action carried additional data — gait, speed, distance, and location — which were analyzed with machine learning. From this rich dataset, the team identified 71 reproducible behavioral sequences that act as the building blocks for more elaborate patterns.
Transitions between these modules imply mechanistic links that produce specific foraging strategies balancing predation risk, energy use and caloric gain. The algorithm also detected spontaneous, individual-specific responses unique to particular mice.
Gregg says the method is sensitive enough to detect the effect of a single-gene mutation. To demonstrate this sensitivity, the team examined mice carrying a mutation in Magel2, an imprinted gene linked to autism-related traits. Imprinted genes are expressed differently depending on whether the maternal or paternal copy is active. Contrary to the assumption that a silent maternal copy would be irrelevant, the analysis revealed measurable behavioral effects from a mutation present only in the mother’s copy.
“What was exciting to us was that we could detect significant behavioral effects from a mutation in only the mother’s gene copy,” Gregg said.
So far, this study focuses on the modular building blocks of foraging in laboratory mice. However, Gregg and colleagues believe the approach could be adapted to dissect other complex behaviors and to identify genomic elements that contribute to human diseases involving behavior, such as obesity, addiction, anxiety and psychiatric disorders.
“By breaking down seemingly spontaneous behaviors into reproducible modules, we uncovered effects that other approaches missed,” Gregg said. “If a human disease is driven by a mutation that alters behavior, this methodology may allow us to map that mutation to specific behavioral modules and better understand how genes shape behavior patterns.”
Authors of the paper, titled Complex Economic Behavior Patterns Are Constructed from Finite, Genetically Controlled Modules of Behavior, include Christopher Gregg, Cornelia Hörndli, Eleanor Wong, Elliott Ferris, Kathleen Bennett, Susan Steinwand, Alexis Rhodes and P. Thomas Fletcher.
Funding: This research was supported by the Swiss National Science Foundation, the National Institutes of Health and the New York Stem Cell Foundation.
Source:
University of Utah Health
Media Contacts:
Julie Kiefer – University of Utah Health
Image Source:
Image credit: Cornelia N. Stacher Hörndli.
Original Research: Open access — “Complex Economic Behavior Patterns Are Constructed from Finite, Genetically Controlled Modules of Behavior”, Cornelia N. Stacher Hörndli et al., Cell Reports. DOI: 10.1016/j.celrep.2019.07.038
Abstract
Complex Economic Behavior Patterns Are Constructed from Finite, Genetically Controlled Modules of Behavior
Highlights
• A robust methodology to dissect the architecture of complex behavior patterns
• Foraging behavior is built from a finite set of genetically controlled modules
• Distinct modules correspond to different economic behavioral strategies
• Parental alleles of the Prader–Willi syndrome gene Magel2 regulate separate modules
Summary
Ethological behaviors can be described as combinations of finite, reproducible modules. Focusing on foraging, the study develops methods to extract these modules from detailed behavioral records in mice. The authors identify discrete modules conserved across strains and ages, alongside sequences that are not modular. Modules vary in form, frequency and timing and appear in probabilistic orders that shape feeding economics, exposure to risk, activity levels and perseveration. The modular architecture changes across development, and developmental stage, genetic background and parental origin each influence specific modules. Dissecting behavior into modules enhances phenotype analysis; notably, both parental alleles of the imprinted gene Magel2 are functional but regulate different modules. Overall, complex economic behavior patterns are constructed from finite, genetically controlled modules.