Virtuelle Methoden für den Entwurf von fokussierten kombinatorischen Bibliotheken basierend auf Feature-Tree-Ähnlichkeit

Combinatorial chemistry allows for the rapid and systematic synthesis of molecules using a unified reaction scheme. The resulting compounds have a common core, but differ in their so-called R-groups. They can be tested for biological activity using high throughput screening (HTS). Due to the large number of possible combinations, only a small number of building blocks can be selected for screening. In particular, for the effective usage of combinatorial chemistry, the building blocks have to be selected in a way, that the resulting products satisfy given criteria. Essential eligibility criteria of the building blocks are the physicochemical properties of the resulting products, for example to increase the chance for bioavailability. Additionally, molecular similarity is a key concept for selection. The resulting products should be similar to known actives and dissimilar to unwanted compounds, for example known inactive molecules. Furthermore, to explore a larger search space, mechanisms are required that lead to more diverse products. In order to fulfill these requirements, the software tool LoFT was developed for the multi-objective optimization of combinatorial libraries. Incorporating the feature tree descriptor, the similarity of the products to one or more query molecules can be calculated without explicitly building them. In combination with product properties, which can be derived directly from the building blocks, an efficient product-based scoring can be obtained on reactant level. Thus, the simultaneous consideration of similarity, diversity and physicochemical properties leads to products, that satisfy the specific needs of the project. Für Katja, Luca und Jona