roshambo is a python package for robust Gaussian molecular shape comparison. It provides efficient and fast algorithms for comparing the shapes of small molecules. The package supports reading input files in the SDF and SMILES formats. It uses PAPER in the backend for overlap optimization.

We recommend that you create a new conda environment before installing roshambo.

conda create -n roshambo python=3.9
conda activate roshambo

roshambo requires a compiled version of rdkit. You need to compile rdkit with the same version of python that you are using to install roshambo. You also need to compile rdkit with the INCHI option enabled. Please refer to the rdkit documentation for installation instructions.

Additionally, since roshambo is GPU-accelerated, you need to have CUDA installed.

After you have installed rdkit and CUDA, you need to set the following environment variables:

export RDBASE=/path/to/your/rdkit/installation
export RDKIT_LIB_DIR=$RDBASE/lib
export RDKIT_INCLUDE_DIR=$RDBASE/Code
export RDKIT_DATA_DIR=$RDBASE/Data
export PYTHONPATH=$PYTHONPATH:$RDBASE

export CUDA_HOME=/path/to/your/cuda/installation

1. Clone the repository:

   git clone https://github.com/rashatwi/roshambo.git

2. Navigate to the roshambo directory:

   cd roshambo

3. Install the package:

   pip3 install .

   Depending on your cluster/machine settings, you might need install in editable mode:

   pip3 install -e .

from roshambo.api import get_similarity_scores

get_similarity_scores(
    ref_file="query.sdf",
    dataset_files_pattern="dataset.sdf",
    ignore_hs=True,
    n_confs=0,
    use_carbon_radii=True,
    color=True,
    sort_by="ComboTanimoto",
    write_to_file=True,
    gpu_id=0,
    working_dir="data/basic_run",
)

The above code will run a similarity calculation between the reference molecule in query.sdf and all the molecules in the dataset.sdf file. Hydrogen atoms will be ignored when aligning the molecules and carbon radii will be used. No conformers will be generated for the dataset molecules. Both shape and color similarity scores will be calculated and the results will be written to a file. The scores will be sorted by the ComboTanimoto score and saved in the directory data/basic_run.

You can also run the above example from the command line:

roshambo --n_confs 0 --ignore_hs --color --sort_by ComboTanimoto --write_to_file --working_dir data/basic_run --gpu_id 0 query.sdf dataset.sdf