Cellular noise

Content sourced from Wikipedia, licensed under CC BY-SA 3.0.

Cellular noise is the random variation in quantities inside cells. Even genetically identical cells can differ in how much protein they produce, their size, or how they respond to signals. Because gene expression helps determine many cell traits, this randomness can affect development, health, and disease.

A common way to measure noise is the coefficient of variation, which is the standard deviation divided by the mean. Other convenient measures used in models include the Fano factor and normalized variance. Researchers study noise mainly in gene expression—the levels of RNA and protein produced by genes—since these influence many cellular properties.

Noise is often split into intrinsic and extrinsic parts. Intrinsic noise comes from random events inside a cell, such as two identical genes expressing at different levels. Extrinsic noise comes from differences between cells, like varying amounts of shared resources or cellular components. Scientists test this with dual-reporter experiments, where two identically regulated genes (often fluorescent markers) are measured in the same cell. If both reporters vary together, it points to extrinsic noise; if they vary independently, intrinsic noise dominates. The interpretation can be subtle; sometimes competition for a limiting regulator can even make the reporters anticorrelated, so deviations from a simple diagonal pattern indicate extrinsic effects. Information theory helps avoid some of these pitfalls.

Because many cellular quantities are discrete (single DNA copies, mRNAs, proteins), researchers use discrete stochastic math to model noise. Core tools include master equations that track the probability of different states over time, often assuming gene activation, transcription, and translation are random (Poisson) processes. These equations can sometimes be solved exactly or approximated with methods like Van Kampen’s expansion. In practice, people run stochastic simulations (Gillespie algorithm) to generate real realizations of cellular processes and estimate statistics.

Inferring model parameters from data is challenging due to sparsity and noise. Researchers use Bayesian methods, such as MCMC and approximate Bayesian computation, which are robust to imperfect data. For specific cases like two-state gene models, moment-based methods can help estimate parameters from mRNA distributions.