📄 systems biology

Sensory variation and behavioural degeneracy: a framework for interpreting heterogeneity in the gut-brain axis

This paper proposes a theoretical framework using an agent-based model to demonstrate that gut microbiome heterogeneity in conditions like autism arises from "behavioural degeneracy," where diverse sensory traits and learning strategies converge on similar dietary patterns and microbial states, thereby challenging the assumption that microbiome differences uniquely reflect intrinsic neurobiological causes.

Original authors: Hunter, W. R.

Published 2026-09-14

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Hunter, W. R.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Scientists have long suspected that the trillions of tiny organisms living in our gut play a significant role in shaping how our brains work and how we behave. This connection, often called the gut-brain axis, has led to a surge in research linking the specific mix of bacteria in our intestines to complex conditions like autism and depression. The prevailing idea has been that if two people have different gut bacteria, they likely have different underlying biology driving those differences. However, a new study suggests this logic might be flawed. It proposes that the path from our internal biology to the bacteria in our gut is not a straight line, but a winding road paved with our daily choices, sensory experiences, and habits. Because our behavior acts as a filter between our brains and our gut, two people with very different internal wiring could end up with nearly identical gut bacteria simply because they made similar food choices.

This complexity is particularly visible in autism, where individuals often experience the world through a unique sensory lens. Some people are hypersensitive to textures or sounds, leading them to avoid certain foods, while others actively seek out specific sensory experiences, driving them to explore new tastes. These sensory differences directly influence what people eat. Since diet is the primary fuel for gut bacteria, these behavioral patterns shape the microbial communities inside us. The problem arises when researchers look at the final result—the bacteria—and try to work backward to find the cause. If two autistic individuals have similar gut bacteria, it is tempting to assume they share the same neurological cause. But if their bacteria are similar because they both happen to eat the same limited range of foods, yet for completely different sensory reasons, then the bacteria alone cannot tell us what is happening in their brains.

To explore this possibility, William Ross Hunter built a computer simulation that mimics how individuals interact with their environment. In this model, virtual agents were given a set of hidden traits that represented how they process sensory information, how much they prefer routine versus novelty, and how sensitive they are to their surroundings. These agents were placed in a simulated world with a variety of food options, each with different sensory characteristics. The agents did not have their diets pre-programmed; instead, they learned what to eat over time based on their hidden traits and their past experiences, much like a person learning to prefer certain flavors through trial and error. As the agents made their food choices, the simulation tracked how these decisions shaped their virtual gut microbiomes.

The results of the simulation revealed a striking pattern of what scientists call degeneracy. In this context, degeneracy means that many different starting points can lead to the same outcome. The study found that agents with vastly different hidden sensory traits often ended up with very similar eating habits and, consequently, very similar gut bacteria. For instance, one agent might avoid a food because it was too intense, while another might avoid the same food because it was too predictable, yet both would end up with a diet that fostered the same type of bacterial community. The simulation showed that these agents naturally organized themselves into three broad behavioral styles: specialists who stuck to a narrow diet, explorers who tried many different foods, and opportunists who fell somewhere in between. Crucially, these styles emerged naturally from the interaction of the agents' traits and their environment, rather than being forced by the model.

The most significant finding was that the link between an agent's hidden traits and its final gut bacteria was weak and non-unique. The simulation demonstrated that you cannot look at a specific gut bacteria profile and confidently say what the underlying sensory traits of the host must be. Different combinations of sensory drive, preference for routine, and environmental sensitivity could all converge on the same microbial state. This suggests that the variability and overlap seen in real-world studies of autism and depression might not be a sign of messy data or poor study design, but rather a fundamental feature of how behavior mediates the relationship between the brain and the gut. The model indicates that similar microbial signatures across different people do not necessarily mean they share the same biological causes; they may simply be the result of different behavioral pathways arriving at the same destination.

This work does not claim to have solved the mystery of the gut-brain axis, nor does it offer a new medical treatment. Instead, it provides a theoretical framework that challenges how scientists interpret their data. By showing that behavior is a powerful, dynamic force that can mask underlying differences, the study argues that researchers need to look beyond simple group comparisons. If the goal is to understand the biological roots of conditions like autism, simply measuring bacteria is not enough. The study suggests that to truly understand the picture, scientists must also measure the sensory experiences and dietary behaviors that sit in the middle, acting as the bridge between the mind and the microbiome. Without accounting for these behavioral pathways, the search for a single biological cause for complex conditions may remain an elusive quest, obscured by the many different ways individuals navigate their world.

Technical Summary: Sensory Variation and Behavioural Degeneracy in the Gut-Brain Axis

Problem Statement
Current research into the gut-brain axis frequently interprets differences in gut microbiome composition as direct reflections of underlying neurobiological differences, particularly in conditions like autism spectrum disorder (ASD) and depression. However, empirical findings are highly variable, with overlapping microbial signatures observed across distinct conditions and within the broad phenotypic spectrum of ASD. This variability complicates causal inference. The core problem identified is that when outcomes are mediated by behaviour—specifically dietary choices driven by sensory processing differences—the mapping from latent biological traits to observable microbial patterns becomes fundamentally non-unique. Existing research often treats behaviour as a secondary covariate rather than a dynamic, context-dependent mechanism, potentially obscuring the true pathways linking neural processes to ecological outcomes.

Methodology
To address this, the author developed a minimal agent-based model (ABM) designed to formalize the pathway linking latent sensory traits, behavioural dynamics, and microbiome outcomes. The model does not impose predefined behavioural categories; instead, behaviour emerges endogenously through the interaction of:

Latent Sensory Traits: Each agent is parameterized by three continuous traits grounded in autism sensory variation: sensory drive (ranging from avoidance to seeking), predictability preference (tolerance for routine vs. novelty), and context sensitivity (modulation by environmental stress and load).
Environmental Context: Simulations occur over discrete time steps with fluctuating environmental variables (stress, sensory load, predictability) that dynamically modulate the agents' latent states.
Reinforcement Learning: Agents select from a structured environment of 100 food items using a utility function that integrates current latent states, food attributes (intensity, novelty, predictability), and learned values ( $Q$ -values). Decisions are probabilistic (softmax) and updated via a standard reinforcement learning rule.
Microbiome Module: A simplified ecological layer maps dietary input to microbial composition. It consists of 120 taxa organized into functional groups. Microbial states evolve dynamically based on dietary choices and a persistence parameter, with outcomes summarized by diversity metrics (richness, dominance, Shannon entropy).

The study utilized 30 independent Monte Carlo simulations with varying random seeds to account for stochasticity. Sensitivity analyses were conducted across ranges of learning rates, choice stochasticity, and microbiome persistence to test the robustness of the findings.

Key Results
The model demonstrated that structured dietary variation and specific microbial outcomes can arise from a minimal set of underlying processes without predefined categories:

Emergent Behavioural Continuum: Agents organized endogenously into a continuous spectrum of strategies: specialists (low dietary entropy, high routine), explorers (high entropy, high diversity), and opportunists (intermediate). These regimes emerged directly from the interaction of latent traits and environmental variability, not from discrete clustering.
Structural Degeneracy: The system exhibited fundamental degeneracy—a many-to-one mapping where distinct combinations of latent sensory traits and behavioural trajectories converged on similar dietary patterns and, subsequently, similar microbiome structures.
Non-Unique Mapping: Pairwise comparisons revealed that while latent distance generally correlated with behavioural distance, the relationship was highly variable. Agents with divergent latent configurations frequently produced low behavioural distances (similar diets), and agents with similar latent configurations produced diverse outcomes.
Microbiome Convergence: Similar microbiome states (in terms of richness and dominance) were generated by distinct behavioural pathways. The model showed that microbiome similarity does not uniquely identify the underlying cause (latent traits), as different sensory profiles can lead to equivalent ecological outcomes through different behavioural routes.
Robustness: The phenomenon of degeneracy persisted across all tested parameter regimes (learning rates, stochasticity, persistence), indicating it is a structural property of behaviourally mediated systems rather than an artifact of specific model parameters.

Key Contributions

Theoretical Framework: The paper provides a conceptual framework for interpreting heterogeneity in gut-brain research, specifically challenging the assumption that microbial similarity implies shared neurobiological mechanisms.
Mechanistic Demonstration: It demonstrates that "degeneracy" in this context arises not merely from stochastic noise but from the structural interaction of state-dependent behaviour, learning, and environmental variability.
Reframing Behaviour: The work positions behaviour not as a nuisance variable to be controlled, but as a mechanistically essential intermediate layer that constrains the interpretability of microbiome data.

Significance and Claims
The paper claims that its primary significance lies in establishing a theoretical limit on causal inference in gut-brain research. It argues that because behaviour acts as a dynamic, context-dependent mediator, observable microbial patterns cannot, on their own, distinguish between shared biological causes and convergent behavioural pathways.

The author explicitly states that the model is a "minimal theoretical demonstration of structural non-identifiability" rather than a quantitatively predictive tool for specific empirical data. The significance is framed as a general principle applicable to any condition where behaviour mediates between neural processes and ecological outcomes (including depression and anxiety), not just autism. The paper concludes that to move beyond group-level associations toward mechanistic understanding, future research must integrate behavioural, sensory, and dietary data alongside microbial profiling, as microbiome data alone is insufficient to resolve the underlying causes of observed patterns.

Technical Summary: Sensory Variation and Behavioural Degeneracy in the Gut-Brain Axis

More like this