Designing new drugs on the computer is one of the main challenges facing computational biophysics. Most of today's structure-based drug design methodologies are based on attempts to fit (geometrically and/or energetically) small molecules into known three-dimensional binding sites ("docking"). Although these methods show much progress they are, unfortunately, relevant only to a small subset of potential drug targets. For most would-be drug targets we do not know the required three-dimensional structure, and thus cannot applied the above methods. In cases where the structure of the target is unknown computational attempts are limited to statistically-based "quantitative structure activity relations" approach or to structural similarity to known ligands.
We present a novel way for structure-based drug design to target molecules who's structures are yet unknown. The new method is based on a comprehensive analysis of the available conformation spaces of different possible drugs and correlating them with known bioactivity and bioselectivity. The idea behind the method is the well known fact that geometric restraining of bioactive peptides can significantly change their bioactivity and/or bioselectivity (even without changing their chemical composition). However, until now there was no method to determine a priory how a given geometric restraints (e.g., cyclization, changing L amino acids to D amino acids) will effect the activity. Using new tools, such as "topological mapping" and "principle coordinate analysis" we have developed a method for quantifying the effect of geometric restraints on the relevant conformation spaces. This quantification together with known data about bioactivity of a few known peptides results in a method with strong predictive power regarding the optimal geometry of the active drug (reflecting the yet unknown structure of the binding site).