You ain't seen nothing, and yet: Future biochemical concentrations can be predicted with surprisingly high accuracy

TL;DR

This work shows that embedding prior knowledge of reproducible, monotone saturating concentration profiles into a Bayesian MAP framework can improve sensing accuracy beyond classic Poisson/Berg–Purcell limits, preserving the $1/\sqrt{N}$ scaling but with a profile-dependent prefactor. By analyzing simple monotone profiles and a piecewise-linear growth model, the authors derive conditions under which the relative error $δc/c$ scales as $1/\sqrt{a^2N}$ with $a>1$, yielding super-Poisson accuracy (e.g., for $a\approx1.8$, ~70% improvement). They also demonstrate that plausible biochemical networks can implement the required MAP computations, enabling on-the-fly prediction of concentration or position in developmental contexts. The results offer a plausible mechanism for the rapid, high-precision fate decisions observed in development and outline how structured spatiotemporal signals can be exploited more broadly in biology. They further show how small stochastic concentration fluctuations around the mean modify but do not overturn the main conclusions, suggesting robustness of the MAP approach to real cellular noise.

Abstract

Accurate sensing of chemical concentrations is essential for numerous biological processes. The accuracy of this sensing, for small numbers of molecules, is limited by shot noise. Corresponding theoretical limits on sensing precision, as a function of sensing duration, have been well-studied in the context of quasi-static and randomly fluctuating concentrations. However, during development and in many other cases, concentration profiles are not random but exhibit predictable spatiotemporal patterns. We propose that leveraging prior knowledge of these structured profiles can improve and accelerate concentration sensing by utilizing information from current molecular binding events to predict future concentrations. By framing the constrained sensing problem as Bayesian inference over an allowed class of spatiotemporal profiles, we derive new theoretical limits on sensing accuracy. Our analysis reveals that maximum a posteriori (MAP) estimation can outperform the classical Berg-Purcell and maximum-likelihood (Poisson counting) limits, achieving a sensing precision of $δc/c = 1/\sqrt{a^2N}$, where $N$ is the number of binding events, and $a > 1$ in certain cases. Thus knowledge of the statistical structure of concentration profiles enhances sensing precision, providing a potential explanation for the rapid yet highly accurate cell fate decisions observed during development.

Figures (3)

• Figure 1: (a) The elliptical shape represents a Drosophila embryo during early development of size $L\approx 500 \mu m$. The grayscale gradient illustrates the steady state concentration profile of the Bicoid protein along the anterior–posterior axis, which is established within approximately one hour after fertilization. Panel (b) shows simulated spatiotemporal evolution of Bicoid gradient. Simulations use a minimal diffusion-degradation model for Bicoid gradient formation driever1988gradient with constant production $j$ at the anterior end, diffusion coefficient $D$, linear degradation with rate $r$, and reflecting boundary condition at the posterior end. This is the simplest model reproducing many features of the experimentally observed concentration profile, although more complex models for better accuracy gregor2007stabilitybergmann2007preabu2010high. Parameter values are $(D, j , r, L) \simeq (5\mu m^2 s^{-1}, 5\mu m^{-2}h^{-1}, 0.5h^{-1}, 500 \mu m)$ and units of concentration is $\mu m ^{-3}$. Panel (c) shows that, in this model, the peak temporal rate of change of the concentration, $\max \partial_t c(x,t)$, scales as a $\sim 1.8$ power of the steady-state level $c_{\infty}(x)$ at the same spatial position.
• Figure 2: Model piecewise-linear concentration profiles defined in Eq. \ref{['eq: piecewise profile with a']} for different values of the parameter $a$. The color of each curve corresponds to its final saturation value $c_{\infty}$.
• Figure 3: Concentration estimation by a biochemical network. (a) Schematic of the biochemical network performing Bayesian inference. Receptor binding event activates the readout molecule $A$. (b) Simulation of the network readout in response to stochastic binding events in a linearly increasing concentration field compared to the approximate MAP and the optimal estimator.