Abstract:Living systems, from single cells to higher vertebrates, receive a continuous stream of non-stationary inputs that they sense, e.g., via cell surface receptors or sensory organs. Integrating these time-varying, multi-sensory, and often noisy information with memory using complex molecular or neuronal networks, they generate a variety of responses beyond simple stimulus-response association, including avoidance behavior, life-long-learning or social interactions. In a broad sense, these processes can be understood as a type of biological computation. Taking as a basis generic features of biological computations, such as real-time responsiveness or robustness and flexibility of the computation, we highlight the limitations of the current attractor-based framework for understanding computations in biological systems. We argue that frameworks based on transient dynamics away from attractors are better suited for the description of computations performed by neuronal and signaling networks. In particular, we discuss how quasi-stable transient dynamics from ghost states that emerge at criticality have a promising potential for developing an integrated framework of computations, that can help us understand how living system actively process information and learn from their continuously changing environment.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the limitations of the current attractor - based framework in explaining computational processes in biological systems. Specifically, the author believes that although the existing attractor theory can explain the stability of certain biological phenomena (such as memory function), it cannot fully capture the ability of biological systems to process dynamic signals, especially the flexibility and adaptability to respond to non - stationary signals in real - time. In order to better understand how biological systems process information and learn from the ever - changing environment, the author proposes the need for a new theoretical framework, namely the framework based on transient dynamics, especially the quasi - stable transient dynamics from ghost states in the critical state. ### Core problems of the paper 1. **Limitations of the existing framework**: - **Attractor theory**: The existing attractor theory mainly focuses on the stable states of the system, which is effective when dealing with static or slowly changing inputs, but is insufficient when dealing with rapidly changing dynamic signals. For example, the bistable system in the cell - signaling network will remain in a stable state after the signal disappears, and thus is insensitive to subsequent signals. - **Dynamic signal processing**: Biological systems need to process multi - sensory, multi - dimensional and often noisy information in real - time in a dynamic environment. The existing attractor theory has difficulty in explaining this real - time processing ability. 2. **Requirement for a new theoretical framework**: - **Transient dynamics**: The author proposes that the framework based on transient dynamics is more suitable as the basis for describing the computational processes in biological systems. Transient dynamics focuses on the transitions between different states of the system, rather than the final stable state. - **Quasi - stable transient dynamics in the critical state**: In particular, the author emphasizes the importance of the quasi - stable transient dynamics (ghost states) that occur in the critical state. These transient states can store information in the system while maintaining the ability to respond to new signals, thereby achieving the "working memory" function. ### Solutions 1. **Ghost States**: - Ghost states are quasi - stable states formed during the transition of the system from one stable state to another. Although these states are unstable, they can temporarily store information, enabling the system to maintain flexibility and adaptability when processing dynamic signals. - For example, in the EGFR network, when the EGF signal disappears, the system will enter a ghost state, which can be maintained for a period of time, thus serving as the cell's memory of the previous signal. This mechanism enables the cell to adapt quickly when encountering new signals. 2. **Hybrid phase - space objects**: - In order to capture more complex natural computational processes, such as sequential responses and context - dependent responses, the author proposes the need to construct hybrid phase - space objects composed of multiple ghost states. These objects can simulate more complex dynamic behaviors and maintain the repeatability of dynamics even under high - noise conditions. ### Conclusion By analyzing the EGFR system in the cell - signaling network, the paper demonstrates the advantages of the new theoretical framework based on transient dynamics and ghost states in explaining the real - time processing of dynamic signals in biological systems. This framework can not only explain the flexibility and adaptability of biological systems, but also provides a new perspective for understanding more complex natural computational processes.

Biological computations: limitations of attractor-based formalisms and the need for transients

Biological computation through recurrence

Limits on the computational expressivity of non-equilibrium biophysical processes

Real-time computing without stable states: a new framework for neural computation based on perturbations

Modeling somatic computation with non-neural bioelectric networks

There's Plenty of Room Right Here: Biological Systems as Evolved, Overloaded, Multi-scale Machines

Attractor and integrator networks in the brain

Biological arrow of time: Emergence of tangled information hierarchies and self-modelling dynamics

Robust Molecular Computation by Active Mechanics

Attractor Stability in Finite Asynchronous Biological System Models

Neural Computations in A Dynamical System with Multiple Time Scales

Form, function, mind: what doesn't compute (and what might)

Bio-inspired computing by nonlinear network dynamics—a brief introduction

Biological Computation as the Revolution of Complex Engineered Systems

Computational Aspects of Feedback in Neural Circuits

Stochastic neurodynamical computation of brain functions

Reply to Comments on Neuroelectrodynamics: Where are the Real Conceptual Pitfalls?

Computational neuroscience: beyond the local circuit

Computational validity: using computation to translate behaviours across species

Nonequilibrium Landscape Theory, Attractor-State Switching, and Autonomous Differentiation

What is Nature-like Computation? A Behavioural Approach and a Notion of Programmability