Abstract : Structural characterization of proteins is often hindered by insufficient amounts of soluble protein. A common approach is to isolate its separate domains, classically done by time-consuming iterations of design, generation and testing of constructs. An alternative approach is to generate a random library of all possible constructs by enzymatic DNA truncation and test them in one experiment for expression and solubility. In this work, the novel, directed evolution-type method Expression of Soluble Proteins by Random Incremental Truncations (ESPRIT) was used to explore the definition of protein domains. The biological focus was a set of multidomain protein kinases that has previously resisted soluble over-expression and structural characterisation. First, improvements in the method were developed to improve the quality and efficiency of the truncation libraries leading to better coverage of the diversity. Then, its application on a more characterised protein, DAPK1, demonstrated precise domain definition. Small variations of the constructs resulted in significantly different thermal stability and crystal packing, thus providing a means to resolve structures with alternative conformations. The method was also used for the identification of soluble variants of the complex between DAPK1 and its partner calmodulin where the minimal interaction region was shorter than previously reported. The application of the method to a series of difficult-to-express protein kinases resulted in rapid identification of incompatibility with E. coli over-expression for some of the targets. While attempts to rescue the low solubility the IKKb kinase catalytic domain were unsuccessful, additional screens on this protein identified soluble and regulatory domains that showed some functionality in vivo. Finally, it was demonstrated that constructs of the PI3Kb p110b catalytic domain with residual solubility in E. coli, identified from screening, could be expressed in insect cells at higher yields in soluble form.