Gene expression at the molecular level. (A) Schematic of a simple stochastic switch whereby a molecule of interest (ie, a gene, mRNA, or protein) undergoes cyclical activation and deactivation reactions, thereby rendering the molecule capable or refractory to an additional reaction. In the example of promoter cycling and transcription, this leads to probabilistically distributed bursts of transcription followed by periods of quiescence. (B) Simulated data showing how for a simple 1-species system produced in bursts and undergoing first-order decay, the probability of firing is directly related to the heterogeneity within a population of identical cells. Histograms on the left show the probability distributions for a gene with a probability of being “ON” (P_on) of 0.1 (black), 0.5 (gray), and 0.8 (red). Plot on the right demonstrates how population noise (captured with the Fano Factor, or the variance/mean) changes as a function of (P_on). A Poissonian distribution (Fano = 1) is the minimal noise observed within biological systems.⁸³  For many TFs, P_on is ∼0.1 to 0.3.⁹³  (C) Relationship of ensemble assays of TF-binding activity to actual behavior of molecules. (i) From the binding distributions derived from ensemble studies such as chromatin immunoprecipitation sequencing, one can conclude that at the green locus, cells in state B immunoprecipitated the locus more frequently than in state A, where states could be defined by a variety of measures including cell surface proteins, cell-cycle position, or metabolic profiling. However, the ensemble studies cannot determine whether this is due to changes in the configurations of chromatin states (ii), each with different affinities for the TF, or from differences in the temporal evolution of the locus (iii), or some combination. Of note, as indicated in the bottom right of iii, state A and state B may lie on a temporal continuum rather than representing discrete entities. This evolution in time is based on energy-dependent modification of the locus by epigenetic enzymes licensed to the site by the TF. Importantly, this implies that cells with identical concentrations of TF could have very different binding patterns (and therefore target gene activity) solely because of each cell’s position on the time curve. This ambiguity obviously complicates the interpretation of ensemble studies of TF binding and how such binding influences target locus expression.