SpaceGA is a tool for accelerated searching of chemical spaces using molecular docking. It integrates the similarity search tool SpaceLight by BiosolveIT with a graph-based genetic algorithm (GB-GA) to efficiently navigate large combinatorial spaces.

SpaceGA explores chemical spaces by using the same crossover module as the GB-GA by Jensen but replaces the mutation module with a similarity search using SpaceLight. This allows SpaceGA to identify "available mutations" within a given chemical space, ensuring that all generated molecules exist within the specified library.

For reproducing the results from the SpaceGA paper, refer to the ./SpaceGApaper directory. The most recent version of SpaceGA is available in the ./SpaceGA directory.

You can install SpaceGA using pip:

pip install -e .

• SpaceLight: A valid license is required to use SpaceGA.
• Lilly Medchem Rules (optional): Supports molecular filtering.
• Additional dependencies are listed in requirements.txt.

You can run SpaceGA directly from the command line using a .json file as input:

python3 /path/to/SpaceGA/main.py /path/to/inputs.json

Alternatively, you can use the provided Jupyter notebooks:

• SpaceGAmanual.ipynb
• SpaceGAautomated.ipynb

These notebooks serve as implementation examples in a Python script.

SpaceGA accepts various arguments specified either in the input .json file or when initializing SpaceGA.

SpaceGA includes two built-in scoring functions that can be specified as follows:

scoring_tool = {"module": "SpaceGA.scoring", "tool": scoringtool}

• FPSearch: Maximizes Tanimoto similarity to a query molecule.
• LogPSearch: Maximizes logP.

No additional arguments are required for LogPSearch.

Users can define and import custom scoring functions following this format:

from abc import ABC, abstractmethod

class Scorer(ABC):
    @abstractmethod
    def __init__(self, arguments):
        pass
    
    def score(self, smi_lst, name_lst):
        pass

Here, arguments is a dictionary containing SpaceGA settings, ensuring seamless integration.

SpaceGA supports filtering based on molecular properties, which can be configured using a dictionary. Below are the available filtering options: