Information Maximization in Biological Signaling and Plasticity

You are reading version 1 of this direction, frozen 2026-10-04. The latest version is version 2.

Version 1 of 236 papers44 conceptsfrozen 2026-10-04

Biological systems are often viewed as messy, stochastic, and constrained by physical noise. However, a growing body of research suggests that from the level of single genes to complex neural circuits, biology behaves like an evolutionarily optimized communication system. By applying the Infomax principle—the idea that a system should maximize the mutual information between its inputs and outputs—researchers have found that biological "wetware" often operates near the theoretical limits of physics.

The Neuron as an Information Channel

In traditional engineering, a communication channel has a fixed capacity. In biology, the channel is dynamic. Martin Stemmler and Christof Koch demonstrated in Information Maximization in Single Neurons that single neurons can maximize information transmission by adjusting their internal ion channel conductances. Instead of simply reacting to stimuli, the neuron "learns" the statistics of its environment and tunes its sensitivity to ensure its limited firing range is used as efficiently as possible.

This tuning extends to the synapses themselves. Péter Adorján and Klaus Obermayer showed in Contrast Adaptation in Simple Cells by Changing the Transmitter Release Probability that by maximizing the entropy of the output distribution, synapses can maintain visual sensitivity across massive ranges of lighting. This suggests that the brain’s primary goal isn't just to "fire," but to ensure that every spike carries the maximum possible amount of "surprise" or information about the world.

Genetic Networks at the Physical Limit

This principle isn't unique to the brain; it appears in the very blueprints of life. Genetic regulatory networks (GRNs) must transmit signals through a soup of molecular noise. Gasper Tkacik, William Bialek, and Aleksandra M. Walczak have shown in a series of papers (e.g., Optimizing information flow in small genetic networks) that these networks are tuned to reach nearly 90% of their theoretical Shannon capacity.

By treating a developing embryo as a communication channel, these researchers found that parameters like protein binding affinities and cooperativity are "set" to maximize the information transferred from molecular inputs to spatial patterns. As Aleksandra M. Walczak notes in her review, biological circuits often operate at the edge of what is physically possible, maximizing precision in the face of unavoidable stochasticity.

Deriving Learning Rules from First Principles

Perhaps the most striking evidence for this idea is that common biological learning rules can be derived mathematically from the goal of information maximization. For decades, Spike-Timing Dependent Plasticity (STDP)—where the timing of spikes determines whether a synapse strengthens or weakens—was seen as a curious biological fact.

However, Gal Chechik and Naftali Tishby in Temporally Dependent Plasticity: An Information Theoretic Account and Taro Toyoizumi and Jean-Pascal Pfister in Spike-timing Dependent Plasticity and Mutual Information Maximization for a Spiking Neuron Model proved that STDP is the optimal strategy for maximizing mutual information between input and output spike trains. In other words, if you were to design a synapse from scratch to be as informative as possible, you would "invent" STDP.

Calculating the Capacity of Life

How do we measure these limits in complex, non-linear systems? The field relies on iterative numerical algorithms to find the input distributions that maximize mutual information.

  1. The Blahut-Arimoto Algorithm: This classic tool is frequently used to calculate the capacity of genetic and molecular channels (Stochastic Proofreading Mechanism Alleviates Crosstalk in Transcriptional Regulation). Recent work by Gabriele Beretta and colleagues has even linked this algorithm to evolutionary game theory, suggesting that the math of information theory and the math of survival are deeply intertwined (Vector Flows and the Capacity of a Discrete Memoryless Channel).
  2. Neural Estimators: For continuous, high-dimensional spaces where traditional algorithms fail, researchers like Dor Tsur and Haim H. Permuter have developed the Directed Information Neural Estimator (DINE). This uses recurrent neural networks to estimate the causal information flow, allowing us to quantify the "intelligence" of a signaling pathway without knowing its underlying probability distributions.

By framing biology as a series of optimization problems—where the objective function is mutual information—we move from describing what biological systems do to understanding why they are built the way they are. Whether it is a neuron in the hippocampus optimizing spatial coding (Michele Nardin) or a protein fold evolving for efficient coding (Abram Magner), the underlying drive is the same: to transmit the most information with the least possible noise.

Papers in this version

History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 25 papers
  2. Version 1 · 2026-10-04 · 36 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_mutual_information_maximization_informat_fe6a9bcb,
  title = {mutual information maximization (information-theoretic synaptic plasticity)},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/mutual-information-maximization-information-theoretic-synaptic-plasticity/rdr_fe6a9bcb1e2f4e9c902ac2b809b38b0f},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 1, frozen text):

@misc{lacuna_information_maximization_in_biological_s_fe6a9bcb_v1,
  title = {Information Maximization in Biological Signaling and Plasticity},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/information-maximization-in-biological-signaling-and-plasticity/rdr_fe6a9bcb1e2f4e9c902ac2b809b38b0f/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}