Strategies Based on Screen Searching

Jakes et al. implemented these screens in a pharmacophoric pattern matching system that is in use at Pfizer Central Research, U.K. ° It allows distances to be specified as either exact or a range of distances and the atoms to be exact matches or any atom type. In the first stage of the search, the bit map of the file of structures is searched by comparison with the bit map of the query screens. Test searches of 10 published pharmacophoric patterns against a subset of the Cambridge Structural Database demonstrate the effectiveness and the efficiency of screen searching. Specifically, the screen search eliminated, on average, over 95% of the molecules in the daubase. 

Sheridan et al. ° use similar screens, except that they distinguish atoms based on aromaticity, hybridization, connectivity, and charge at neutral pH. They also calculate screens based on the positions of the lone pairs on heteroatoms and of the center of mass of rings. Thus, they have many more screens. To accommodate this, they use an inverted bit map, that is, one in which each row contains the information on which molecules possess a particular screen. This facilitates searching for example, screening 223,000 structures typically requires about 1 min CPU time on a VAX 8650. However, updating an inverted bit map when a new compound is added is much more demanding of computer resources. 

In the systems at Pfizer, the molecules that pass the screens are carried onto the distance search and the geometric search. This is necessary since a struaure that matches the query at the screen level may not be a match, either because the screens refer to distance ranges and not to precise distances, or 

For the distance search, the coordinates are retrieved from disk the exaa distances are then calculated and compared with the set of query distances. In addition to the added time for disk access, the calculation of the distances and the comparison with the set of query distances also require computer time. Therefore, the distance search is much more time consuming than the saeen search, despite the much smaller number of structures processed. 

A recently released three-dimensional structure searching system is MACCS-3D, which is designed to be coupled with a MACCS-II database enhanced with three-dimensional structural data. The program uses thousands 

---