The specific labeling of proteins with synthetic probes is a powerful approach to study protein function and protein tags have been widely used for this purpose. A well-established example for a self-labeling protein tag is SNAP-tag. It specifically reacts with a wide variety of O6-benzylguanine derivatives (BG-derivatives) and was derived from the human O6-alkylguanine-DNA alkyltransferase (hAGT) by protein engineering. Relative to hAGT, SNAP-tag possesses a 52-fold higher reactivity towards BG-derivatives, does not bind to DNA and expresses well in cells as on cell surfaces. It is known that alkylation of hAGT results in protein unfolding and degradation. However, an increased degradation of SNAP-tag fusion proteins after labeling has not been observed. The first part of this work focused on the structural basis underlying the differences in protein stability between SNAP-tag and hAGT. A detailed biochemical and structural analysis was performed to determine (i) the interaction of SNAP-tag with its substrate, (ii) the factors responsible for its increased reactivity and (iii) how the labeling affected the stability of the protein. Besides an increased reactivity with BG-derivatives the superior stability of SNAP-tag compared to the parent protein hAGT could be confirmed. Whereas wild-type hAGT was rapidly degraded in cells after alkyl transfer, benzylated SNAP-tag showed a higher stability against proteolytic degradation. Moreover, the combination of our crystallographic and computational data provided further insight into the structural basis for the improved properties. The data indicated that the intrinsic stability of a key alpha helix was an important factor in triggering the unfolding and degradation of wild-type hAGT and provided new insights into the structure-function relationship of this DNA repair protein. The second part was aiming for the generation of a new SNAP-tag-based inhibitor complex. It was envisaged that this complex would interact with the target protein via amino acid loops and decrease its function only upon labeling with BG-inhibitor molecules. Therefore, SNAP-tag was modified by the insertion of stretches of randomized amino acids and the generated protein libraries were screened for binding affinity. The utilization of two yeast-based systems, the yeast three-hybrid and two hybrid technologies, allowed for the differentiation of small-molecule dependent and independent binding interactions. It could be demonstrated that a specific protein-loop interaction can be generated by this approach. It could further be shown that inhibition of the catalytic activity of the target protein E.coli dihydrofolate reductase by a SNAP-loop mutant was possible. In summary this work revealed new insights into the stability of hAGT and SNAP-tag and the structure-function relationship of AGTs in general. Further, SNAP-tag engineering generated a new protein-binder whose affinity towards the target protein was leading to protein inhibition.