It is well documented that molecular processes can be thermodynamically coupled such that the shift in the equilibrium of one process (e.g. ligand binding) can modify the kinetics and/or equilibrium of another process (e.g. receptor activation). This form of thermodynamic coupling is known as allostery and is believed to be a ubiquitous mechanism of function throughout cell, especially in the function of membrane proteins such as G protein-coupled receptors and transporters. In addition, the existence of ligand-specific allosteric modulation in both transporters and GPCRs emphasizes the importance of understanding how allostery works in these systems in terms of atomic-level physical mechanisms. Towards that goal, the work described in this dissertation will focus on two specific aims: i) the development of theoretical models that provide insight into the structural and dynamic features required for systems to be allosteric, and ii) the development of computational methods that can identify these features in specific systems of interest. First, we present a new theoretical model of allostery, the Allosteric Ising Model, which leads to several analytical conclusions regarding the structural and energetic requirements for long-distance allostery. Next, we present N-body Information Theory (NbIT) analysis, which improves on existing methods for identifying the structural components that act as allosteric channels. We illustrate the power of NbIT by identifying the allosteric channel underlying allosteric modulation of intracellular domain motions by substrate in LeuT. Then we present a random forest-based method for identifying class-specific behavior from ensembles of the same protein bound to different ligands. This method is able to identify interactions that respond in a hallucinogen-specific manner in the serotonin receptor 5-HT2AR. Finally, we present a generalized form of the two-state allosteric efficacy that can be applied to discrete and continuous variables. This description of allosteric coupling suggests that mutual information, a common measure of allostery, is fundamentally related to allostery but in itself is not a good quanitifcation of it. The new quantification of allosteric coupling is then used to identify allosteric couplings in the simplest allosteric system, alanine dipeptide.