Abstract

The binding of insulin to the extracellular part of the insulin receptor is a key step in the insulin signalling pathway. Upon binding, the receptor is autophosphorylated and the intracellular tyrosine kinase is activated. In 1999, Zhang et al. published a small molecule identified from a fungal extract, which activates the human insulin receptor by binding directly to the intracellular domain of its beta-subunit. This compound (demethylasterriquinone B-1, DMAQ-B1) was shown to lower blood glucose levels in mouse models of type 2 diabetes mellitus. During the last years, structures and activities of approximately 100 derivatives of this compound have been published. Most of these structures contained a quinone substructure, which might cause toxic side effects. Since treatment of type 2 diabetes includes long-term administration of anti-diabetic compounds, it would be beneficial to find compounds with a different type of structure which activate the insulin receptor. The aim of this dissertation was to build computational models which can be used to screen for new insulin-mimetic compounds and subsequent validation of the models by testing some of the obtained hits in relevant biological (i.e. cell-based) experiments. Three different ligand based computational methods, namely self-organizing maps, fingerprint similarity and shape similarity, have been used to screen a large vendor database for potential insulin receptor activating compounds. By testing 13 representative compounds from the identified scaffolds we found three compounds which are able to activate Akt kinase, an important downstream target of the activated insulin receptor. One of the compounds increased glucose uptake in muscle cells. Derivatives of these compounds were further investigated to gain information on structure activity relationships. Additionally, the toxicity of the compounds in cells was assessed to show that the insulin-mimetic activity of our identified molecules is not correlated with toxic effects.