[Paper Review] Optimal information gain at the onset of habituation to repeated stimuli

Giorgio Nicoletti, Bruzzone, Matteo+3 more|arXiv (Cornell University)|Jan 30, 2023

Neural dynamics and brain function4 citations

TL;DR

This paper proposes a thermodynamic framework linking adaptation in biological systems to optimal information processing and minimal energy dissipation. By modeling receptors, feedback, and slow information storage, the authors show that adaptation emerges naturally in a narrow regime maximizing information gain while minimizing dissipation—verified in zebrafish neural responses to repeated visual stimuli.

ABSTRACT

Biological and living systems process information across spatiotemporal scales, exhibiting the hallmark ability to constantly modulate their behavior to ever-changing and complex environments. In the presence of repeated stimuli, a distinctive response is the progressive reduction of the activity at both sensory and molecular levels, known as habituation. In this work, we solve a minimal microscopic model devoid of biological details, where habituation to an external signal is driven by negative feedback provided by a slow storage mechanism. We show that our model recapitulates the main features of habituation, such as spontaneous recovery, potentiation, subliminal accumulation, and input sensitivity. Crucially, our approach enables a complete characterization of the stochastic dynamics, allowing us to compute how much information the system encodes on the input signal. We find that an intermediate level of habituation is associated with a steep increase in information. In particular, we are able to characterize this region of maximal information gain in terms of an optimal trade-off between information and energy consumption. We test our dynamical predictions against experimentally recorded neural responses in a zebrafish larva subjected to repeated looming stimulations, showing that our model captures the main components of the observed neural habituation. Our work makes a fundamental step towards uncovering the functional mechanisms that shape habituation in biological systems from an information-theoretic and thermodynamic perspective.

Motivation & Objective

• To identify the universal principles underlying biological adaptation across spatiotemporal scales.
• To resolve the gap between microscopic biochemical mechanisms and macroscopic adaptive behaviors.
• To establish that adaptation arises from a trade-off between information gain and energy dissipation.
• To demonstrate that negative feedback and slow information storage are necessary and sufficient for adaptive responses.
• To validate the framework using large-scale neural data from zebrafish larvae under repeated visual stimulation.

Proposed method

• Formalizing adaptation as a stochastic process with receptor states (active/passive), readout populations (U), and storage molecules (S) with time-scale separation.
• Deriving the internal entropy production rate $\dot{\Sigma}_{\text{int}}$ using master equation dynamics and joint probability distributions.
• Applying the timescale separation approximation to reduce the full dynamics to a slow storage variable governed by feedback.
• Using a minimal model where each readout unit represents a neural population of 100 neurons firing stochastically with probability 0.5.
• Fitting experimental zebrafish neural activity data with a linear model using stimulus-convolved kernels to extract responsive neurons.
• Reconstructing raster plots from the model to compare low-dimensional adaptation dynamics with experimental data.

Experimental results

Research questions

• RQ1 What thermodynamic principles govern the emergence of adaptive responses in biological systems?
• RQ2 How does the trade-off between information gain and energy dissipation shape the dynamics of adaptation?
• RQ3 Which minimal components (feedback, storage, out-of-equilibrium dynamics) are necessary and sufficient for adaptation?
• RQ4 Can a generic model based on information-thermodynamic principles reproduce experimental neural adaptation patterns?
• RQ5 To what extent is adaptation independent of biological details and universal across scales?

Key findings

• Adaptation emerges in a narrow optimal regime where information gain is maximized and internal dissipation is minimized.
• The system reduces entropy production over time, indicating a progressive decrease in energy waste during adaptation.
• The efficacy of feedback mechanisms increases during adaptation, correlating with reduced dissipation.
• Negative feedback and a slow information storage mechanism are necessary and sufficient for adaptation, independent of biological specifics.
• The model successfully reproduces the low-dimensional adaptation dynamics observed in zebrafish larval neural responses to repeated visual stimuli.
• Experimental data from ~2400 responsive neurons in zebrafish larvae show a decay in response amplitude over repeated stimuli, consistent with the model’s predictions.