Split-protein sensors have become an important tool for the analysis of protein-protein interactions in living cells. In general, two interacting proteins are expressed as fusion proteins with a pair of inactive fragments of a reporter enzyme. Interaction-induced reassembly of the two fragments then results in a functional enzyme and a detectable phenotypic readout. Despite the constantly expanding repertoire of methods, new split-protein sensors that could detect and screen for protein-protein interactions both in the cytosol and in the membrane would be very useful. In a first attempt to create new split-protein sensors, cytochrome c peroxidase (CCP) from the yeast Saccharomyces cerevisiae was rationally dissected into two fragments, which were fused to two interacting proteins. Activity of reassembled peroxidase might be visually detected using a simple colony screen [1]. However, this approach failed due to the insolubility of the chosen fragments. A random approach based on circular permutation originally developed by Graf and Schachmann was therefore adapted to isolate suitable fragmentation sites [2]. Unfortunately, only split-proteins expressing quasi wild-type protein were isolated, resulting from cuts close to the N or the C terminus of CCP. Two reasons can account for this: (i) the fragile active site environment of CCP could considerably hamper the fragmentation of the enzyme; (ii) the method itself favors the isolation of quasi wild-type proteins. The combinatorial method for the generation of new split-protein sensors was therefore further modified to circumvent the isolation of quasi wild-type proteins and successfully applied to the (β/α)8-barrel enzyme N-(5'-phosphoribosyl)-anthranilate isomerase Trp1p from Saccharomyces cerevisiae. The generated split-Trp protein sensors allow for the detection of protein-protein interactions in the cytosol as well as the membrane by enabling trp1 cells to grow on medium lacking tryptophan. In addition, split-Trp can be used as reporter for the detection of small molecule-protein interactions. This powerful selection thus complements the repertoire of the currently used split-protein sensors and provides a new tool for high-throughput interaction screening. Furthermore, the combinatorial approach should be able to generate split-protein sensors of almost any protein, thereby yielding tailor-made sensors for different applications.