Computer-Aided methods as tools to find new bioactive compounds
Project leader: professor ANTTI POSO
Department of Pharmaceutical Chemistry, University of Kuopio
Doctoral students of the project:
Tuomo Kalliokoski
Other researchers of the project:
Maija Lahtela-Kakkonen
Key words: molecular modeling, virtual screening, software development, structure-activity relationships, bioinformatics
Project desciption and main results:
Whenever a new interesting protein is found the next question is usually: "what molecules can interact with this protein". Even if we do not know the 3D-structure of the protein, comparative modeling can solve the 3D structure of a target. In principle this gives us a possibility to use target-based drug design methods, like virtual screening and molecular docking to search new bioactive compounds.
Unfortunately the utilization of these computer-based methods has been unsatisfactory. Still too often specialists are working apart each other; research groups doing biochemical or drug synthesis are not specialist in bioinformatics and modeling. This is quite natural since most of the modeling methods are only used within short period of time during an individual research project. As an example virtual screening methods are only used in the beginning of individual research project: once the scientist has learned (usually with trial and error) the skill of virtual screening he/she will not need that skill any more.
The general aim is to continue the development of research group in computer-aided, target based drug design. The knowledge of group is utilized in ongoing drug research and development projects at the Department of Pharmaceutical Chemistry, KU. In this proposal in silico screening and Computer-Aided Drug Design (CADD) are used to find new bioactive compounds. The methods to be used are both currently available CADD methods, like molecular docking, database screening, 3D-QSAR, comparative modeling and some new methods based on inventions made in KU.
Although in each case there are unique problems the overall research path is usually the following. At the first stage the model of the target protein is constructed, either using comparative modeling tools, or preferably crystal structure. At the same time possible structure-activity data is collected (either from literature or from collaborators experiments) and 3D-QSAR (usually CoMFA or CoMSIA models) models are created. Using this data a simple search query is constructed. Typically the first search results 104-105 compounds. These compounds are further processed using ADMET-type filters, starting from simple Lipinski-rule filters but also different 3D-molecular field based filters or fragment-based ADMET filters may be used. In addition new lead compounds are found using molecular field based database search methods developed by Antti Poso and his research group (method is not public). At the next step compounds are virtually docked into the binding cavity of the protein(model). Currently available docking tools, like Gold and FlexX are powerful enough to allow dockings of thousands of molecules within reasonable timeframe and accuracy. Dockings are analyzed using different scoring functions, especially with consensus scoring approach, and a selection of compounds are actually purchased (or synthesized) and tested. Selection is based not only on docking results, but also synthetic feasibility and known 3D-QSAR models can be used as additional tools to find best possible candidates. The feedback from these tests will readjust the models and search methods, and also give valuable information concerning the performance of used methods. This data can be used to find better compounds and to create better software. Drug synthesis work for lead optimization is further guided by modeling results, especially docking and 3D-QSAR methods are critical in this phase.
Publications: