It is commonly assumed that a protein exists in a single structure and that evolution molds this single structure to perform a chosen function. However, physical chemistry suggests that proteins can adopt numerous conformations. The serendipitous discovery of cryptic allosteric sites—pockets in a folded protein that are invisible to conventional experiments but which can alter enzymatic activity via long-distance communication—supports the existence of a surprising variety of conformations. The ability to identify all the conformations a protein visits would greatly enhance our understanding of protein function and open new possibilities for drug design, such as the ability to intentionally target cryptic allosteric sites. In this talk, I introduce a computational approach, called Markov state models, which I developed to quantitatively describe the conformational space a protein explores. Using these models, I show that a known cryptic site in TEM-1 β-lactamase can be identified from the equilibrium dynamics of this protein even in the absence of any ligands. Moreover, I demonstrate that my models predict the existence of numerous additional cryptic allosteric sites. Inspired by my predictions, I have conducted experiments testing the presence of these new cryptic sites. The results of these thiol-exchange experiments confirm the existence of multiple allosteric binding pockets. Application of these methods to other proteins would facilitate drug design by greatly expanding the repertoire of druggable sites. Furthermore, these result open exciting opportunities for exploring other ways in which evolution has made use of protein flexibility.