However, at present no consensus exists regarding the basal level of variability inherent to steady-state gene transcription, and we expect that the magnitude of this noise would vary with absolute mRNA quantity and depend upon the intrinsic biochemical properties of specific genes or gene classes. Given the current limitations in measurement technology, such a dynamical systems approach to characterize baseline transcriptional heterogeneity becomes unwieldy for even very small numbers of genes, suggesting that an absolute threshold for homogeneity will be difficult to define. Alternatively, one could apply traditional statistical methods to compare the variability observed across a given population against that of a ����control���� group, evaluated using an identical panel of genes. However, the multipotent nature of LT-HSCs is such that those genes which best characterize this population are not, to our knowledge, universally expressed across any other cell type. Further, the capacity of LT-HSCs for differentiation has precluded comparative evaluation of a clonal KRM-III LT-HSC population. These inherent limitations are not unique to LT-HSCs, and may be relevant to the study of many rare cell populations. These factors have motivated us to develop an approach using principles of Indatraline hydrochloride information theory and statistical physics to test the hypothesis of relative transcriptional homogeneity. Information theory focuses on understanding and correcting for randomness or entropy within a dataset to allow quantification and interpretation of heterogeneous data, and work in statistical physics has generated methods for applying probability functions to inherently stochastic processes. In the absence of an acceptable external comparison, these methods permit us to utilize relationships derived from the variability within our data itself in order to provide insight into the dynamics of this complex system. This approach itself is not novel, and similar methods have been applied with great success to problems in signal processing and control theory ; however, these techniques have only recently gained traction as tools to characterize biological systems. In order to determine whether LT-HSCs represent a homogeneous population or several discrete subpopulations, we applied a unifying procedure for model selection and multimodal inference based on the principles of information divergence, originally described by Kullback and Leibler.