Skip to main content
Help the Python Software Foundation raise $60,000 USD by December 31st!  Building the PSF Q4 Fundraiser

Guacamol: benchmarks for de novo molecular design

Project description


Build Status

GuacaMol is an open source Python package for benchmarking of models for de novo molecular design.

For an in-depth explanation of the types of benchmarks and baseline scores, please consult our paper Benchmarking Models for De Novo Molecular Design


The easiest way to install guacamol is with pip:

pip install guacamol

guacamol requires the RDKit library (version 2018.09.1.0 or newer).

Benchmarking models

For the distribution-learning benchmarks, specialize DistributionMatchingGenerator (from guacamol.distribution_matching_generator) for your model. Instances of this class must be able to generate molecules similar to the training set.
For the actual benchmarks, call assess_distribution_learning (from guacamol.assess_distribution_learning) with an instance of your class. You must also provide the location of the training set file (See section "Data" below).

For the goal-directed benchmarks, specialize GoalDirectedGenerator (from guacamol.goal_directed_generator) for your model. Instances of this class must be able to generate a specified number of molecules that achieve high scores for a given scoring function.
For the actual benchmarks, call assess_goal_directed_generation (from guacamol.assess_goal_directed_generation) with an instance of your class.

Example implementations for baseline methods are available from


For fairness in the evaluation of the benchmarks and comparability of the results, you should use a training set containing molecules from the ChEMBL dataset. Follow the procedure described below to get standardized datasets.


You can download pre-built datasets here:

md5 05ad85d871958a05c02ab51a4fde8530 training
md5 e53db4bff7dc4784123ae6df72e3b1f0 validation
md5 677b757ccec4809febd83850b43e1616 test
md5 7d45bc95c33c10cb96ef5e78c38ac0b6 all


To generate the training data yourself, run

python -m -o [output_directory]

which will download and process ChEMBL for you in your current folder.

This script will use the molecules from holdout_set_gcm_v1.smiles as a holdout set, and will exclude molecules very similar to these.

Different versions of your Python packages may lead to differences in the generated dataset, which will cause the script to fail. See the section below ("Docker") to reproducibly generate the standardized dataset with the hashes given above.


To be sure that you have the right dependencies you can build a Docker image, run from the top-level directory:

docker build -t guacamol-deps -f dockers/Dockerfile .

Then you can run:

docker run --rm -it  -v `pwd`:/guacamol -w /guacamol guacamol-deps python -m -o guacamol/data



Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for guacamol, version 0.5.3
Filename, size File type Python version Upload date Hashes
Filename, size guacamol-0.5.3-py3-none-any.whl (38.0 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size guacamol-0.5.3.tar.gz (30.4 kB) File type Source Python version None Upload date Hashes View

Supported by

Pingdom Pingdom Monitoring Google Google Object Storage and Download Analytics Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page