De novo design and synthetic accessibility

The design and development of a new drug is an extremely costly and time-consuming process. Therefore, in recent decades novel experimental and computational techniques have been introduced to make this process more efficient. Bioinformatics and chemoinformatics methods have become an integral part of the drug design progress. At the outset of the design of a new drug stands the quest for a novel structure that may carry the desired biological property. Both structureand ligand-based methods have been developed to find new lead structures. Structure-bases methods need knowledge on the three-dimensional structure of the target protein. Then, either a de novo design method or a docking procedure can be applied. In de novo design, molecules are grown into the binding pocket of the protein so as to fill this pocket and provide sites for binding the ligand to the protein. In a docking experiment, threedimensional structures are taken from a database and are fitted into the binding pocket so as, again, to fill this pocket and satisfy sites for binding. Ligand-based methods do not need knowledge on the three-dimensional structure of a protein; in fact, they even do not need knowledge on what protein is actually involved for targeting. Rather, ligand-based methods try to find new lead structures by deriving knowledge from known ligands to this protein and use this knowledge to search in a database of structures, either by pharmacophore or by similarity searching. De novo design methods can look back on a long history; the first systems were developed by Jeff Howe at Upjohns (Moon) and by Hans-Joachim Bohm at BASF (Ludi). Since then, several other systems have been designed. One of the drawbacks of de novo design systems is that they may generate a large number of structures, some of them quite complicated and therefore difficult to synthesize by requiring long-step syntheses with carefully controlled reactions, in particular as stereochemistry is concerned. It should also be mentioned that docking procedures and ligand-based methods might run into the same kind of problems if large databases containing virtual structures are used. As, eventually, these novel structures have to be synthesized and tested, a costly process, it becomes imperative to rank these structures according to their ease of synthesis, according to synthetic accessibility, in order to minimize the efforts in the laboratory. If a reasonable ranking is found that is also accepted by medicinal chemists trust into computational methods can be generated. Thus, chemoinformaticians and medicinal chemists will work closely together, a process that is so essential for success in drug design. The American Chemical Society and the Chemical Structure Association have recognized the importance of de novo systems and the crucial importance of estimating synthetic accessibility and have asked me to organize a symposium on ‘‘de novo Design and Synthetic Accessibility’’ at the American Chemical Society Spring Meeting in Atlanta, GA, in March 2006. Unfortunately, not all major players in these fields could be gained to participate in this symposium but the program could provide a good overview of the state of the art. The program consisted of eight presentations: