GoForth Algorithm

• Read a file of SMIRKS (transforms).

• Read an initial population of N[generation=0] SMILES.

• Create a Thor database. Load initial SMILES.

• FOR GENERATION (I = 1...) LOOP until a QUIT CONDITION:
  • For each molecule in database of generation (I-1), apply every transform, yielding a set of reactions and products.
  • For each successful transform, examine products (total M[I]) for FITNESS CONDITIONS.
  • (Optional) Sample fit products.
  • If not already found, and if FITNESS/sampling conditions are met, write product SMILES to database.
  • Report new population size (N[I]).

• FITNESS CONDITIONS (examples):
  1. Valid ClogP calculated.
  2. ClogP value requirement.
  3. Reject SMARTS matches with -badsmarts file.
  4. C:O ratio requirement.
  5. Gibbs free energy < ? (not implemented)

• QUIT CONDITIONS:
  1. population converges
  2. population limit reached
  3. generation limit reached

I = generation
N[I] = size of unique-SMILES population
M[I] = number of generation-I products