Requirements

• Run poetry install to install all the packages needed.

• Get the guacamol smiles with bash guacamol_baselines/fetch_guacamol_dataset.sh

• Unzip de folder models.zip in synthetic_scorers/RAscore

• During the generation, the smiles sampled are stored in Mongo database, so you need to have two environment variables : MONGO_URL with the correct url, and DB_STORAGE with the name of the database.

• If you want to use the score RScore, given by Spaya API, you need to have two more environment variables : SPAYA_API_URL and SPAYA_API_TOKEN, which are your credentials to use the Spaya API.

Package description

• This package is a fork of the code of BenevolantAI https://github.com/BenevolentAI/guacamol_baselines. It enables running generations with several models, for the tasks from guacamol benchmark , and for another task named "pi3kmtor" described in the paper.
• Additionaly, the package enables to run generations with synthetic score as a new constraint. The synthetic accessibility score can be one of the following :
  • SA score
  • RA score
  • SC score
  • RScore (SpayaApi)
  • RSPred (Predictor of SpayaApi trained on Chembl24).

Run generations

The goal directed generations have 2 essentials arguments :

• suite : this value will determine the generations that are going to be launched. A value of a suite correponds to a list of tasks.

  • 'guacamol_paper' launches the generation of guacamol benchamrk tasks with successively no synthetic scores, SA score, RScore and RSPred.
  • 'pi3kmtor_paper' launches the pi3kmtor generation with successively each of the synthetic scorers.
  • 'pi3kmtor' just launches one generation of pi3kmtor, with the synthetic score indicated in the variable "synth_score"

• synth_score : when indicated, tells the synthetic scorer to add in the generation, should be one of RScore, SAscore, SCscore, RAscore, RSPred. There is not need to use synth_score if suite is 'guacamol_paper' or 'pi3kmtor_paper'.

For instance To run 10 steps of generation around pi3kmtor dataset, optimizing 4 constraints and with SA score constraint run:

poetry run python -m guacamol_baselines.smiles_lstm_hc.goal_directed_generation --suite pi3kmtor --n_epochs 10 --synth_score SAscore

Get the results

• The molecules generated are in the mongo database in your var env DB_STORAGE, with in the collection under the name ``benchmark_name+"_"+synth_score```

• For the pi3kmtor generations, the differents scores of the molecules are already in the collection.

• nYou can use the notebook in exploit_results/exploit_results_pi3kmtor.ipynb to analyse the results.