GADD Example Project: Results

Performance:
Feb. 14: 10 hrs, 4 gens, 485k UMOLs, 100k FMOLs (195MHz R10K SGI)

Questions of interest:

How druglike is the GADD database?
How similar to Medchem00 is the GADD database?
How random and diverse is the GADD database?

One basic measure of how similar the GADD database is to the original seed database is how many smiles are in common.

$ thorcommonsmi gadd medchem00 RESULTS(thorcommonsmi): 877 smiles shared by: gadd@origin 112,866 (100,000 FMOLs + fragments) medchem00@origin 36,398
Implication: GADD is building Medchem-like molecules. GADD is building new molecules.

Repeat experiment:
$ thorcommonsmi gadd gadd2 RESULTS(thorcommonsmi): 14850 smiles shared by: gadd@origin 112,866 (100,000 FMOLs + fragments) gadd2@origin 112,866 (100,000 FMOLs + fragments)
Implication: GADD is plenty random.

Further evaluation of results is desireable but beyond the current scope of this project.

MUG'01 -- 6-9 March 2001 -- Santa Fe, NM

Daylight Chemical Information Systems Inc.
info@daylight.com