MODSMI transformations

The atoms on which the MODSMI transformations are to be applied are identified by a GENIE search that recognises atoms or bonds from their substructural environment. They are highlighted for ALADDIN processing with the GCL command PRINT /BOLDID/ . Because the hit atoms from an ALADDIN geometric test are labelled with the geometric object to which they correspond, MODSMI transformations can be applied to all atoms that are in a particular substructural environment or to only such atoms with a particular... 

The simplest MODSMI transformation made to the identified atom or group of atoms is NIBBLE. It removes the identified atom(s) from the SMILES string. Thus NIBBLE called iteratively will transform a pendant propyl into a pendant ethyl into a pendant methyl into no pendant group. For example NIBBLE was used to remove the OMe and Me groups of structure 1 as part of the transformation of it into structures 7 and 11. 

The command HALT indicates that processing on that SMILES is to stop, the programmed MODSMI transformation is to be made, and re-execution of the entire ACL using the newly modified smiles is to be performed. The MODSMI transformations are performed iteratively until no more can be apphed. By appropriate nesting very elaborate transformations can be made to a 2-D structure. 

One can use standard GCL commands to print any current structure, intermediate or final in the MODSMI transformation, to an output file. However, an ACL command to stop modification of the current SMILES has also been added HALT LOGSMI QUIT . When this statement is executed, the current SMILES is output into a file with a name derived from the name of the input ACL file plus an... 

Tjrpically these MODSMI transformations are performed on SMILES previously identified by an ALADDIN search. However they may be apphed to molecules identified by any procedure. When MODSMI is used the molecules are not in a database, but are presented as an input file. 

The program ONESMILE removes duplicate structures from a sorted list of SMILES. Thus, after the MODSMI transformations, sorting and ONESMILE would be used to produce a file of the unique molecules. Notice that this is possible only because all MODSMI operations produce the unique SMILES with the result that each particular molecular structure is represented by the same SMILES string regardless of the order of the atoms in the structure from which it originated. 

The ACL shown in Figures 5 and 7 was used to transform the database structures into molecules to be synthesised. The ALADDIN search of a database of carefully modelled 3-D structures used in studies of a variety of medicinal chemistry projects identified 13016 potential compounds for synthesis. The MODSMI transformations reduced this number to 2529, an 80% reduction. Two slightly different searches of CONCORD structures of compounds in the Fine Chemicals Directory produced 12996 and 6559 original hits that were transformed into 1874 and 2247 unique compounds for an 85% and 66% reduction, respectively. Finally, the ALADDIN search of CONCORD structures of compounds in POMONA89 resulted in 10319 hits which were transformed into 2314 unique compounds for a 78% reduction in the number of compounds to be considered. Clearly the MODSMI transformations reduce the number of compounds to be handled by later steps. Thus this is a very useful strategy for evaluation of hits from a 3-D search. 

Of these, 259, 124, 76, and 160 compounds respectively match the total search target. The low percentage of designed compounds that meet the total criteria was a source of concern. However, it was shown not to be due to the MODSMI transformations but rather because firstly, CONCORD generates only one conformation that might be different from that of the database molecule and secondly, the original search used broad criteria in order not to miss any interesting molecules. Each of these is responsible for approximately 50% of the failures . 

For bond transformations, GrCL is used to identify the atoms to be present at each end of the bond. MODSMI provides two elementary transformations to be made to bonds. JOIN connects unconnected atoms. AXE n changes the bond order of an existing bond by n, either positive or negative. Thus AXE -1 would break a single bond. 

By combining MODSMI, ONESMILE and ATOMWT operations, one can transform a list of ALADDIN hits into a much smaller list that preserves the... 

---