This repository is the official implementation of the paper “Limeade: Let integer molecular encoding aid”. Please cite as:
- Shiqiang Zhang, Christian W. Feldmann, Frederik Sandfort, Miriam Mathea, Juan S. Campos, Ruth Misener. "Limeade: Let integer molecular encoding aid." arXiv preprint arXiv:2411.16623 (2024).
The BibTex reference is:
@article{zhang2024limeade,
title = {Limeade: Let integer molecular encoding aid},
author= {Shiqiang Zhang and Christian W. Feldmann and Frederik Sandfort and Miriam Mathea and Juan S. Campos and Ruth Misener},
journal = {arXiv preprint arXiv:2411.16623},
year = {2024},
}
Most functionalities are demonstrated using Jupyter notebooks available in the notebooks folder.
To install requirements:
pip install -r requirements.txt
Limeade relies on Gurobi to generate feasible solutions as default. A license is needed to use Gurobi. Please follow the instructions to obtain a free academic license. Limeade also proves a Pyomo version (with CPLEX as default) so that the users could use open-sourced solvers.
To generate molecules with N atoms choosing from atom list atoms, run this command (use N=10 and atoms=["C", "N", "O", "S"] as an example):
from limeade import MIPMol
Mol = MIPMol(atoms=["C", "N", "O", "S"], N_atoms=10)
To set lower and upper bounds for each type of atom, run this command:
Mol.bounds_atoms(lb, ub)
where lb (and ub) is a list with length equal to atoms giving the minimal (and maximal) number of each type of atom.
Similarly, to set bounds for number of double/triple bounds and rings, run these commands:
Mol.bounds_double_bonds(lb_db, ub_db)
Mol.bounds_triple_bonds(lb_tb, ub_tb)
Mol.bounds_rings(lb_r, ub_r)
where lb_db (and ub_db) is the minimal (and maximal) number of double bonds, lb_tb (and ub_tb) is the minimal (and maximal) number of triple bonds, lb_r (and ub_r) is the minimal (and maximal) number of double rings.
To include a given list of SMARTS strings substructures, run this command:
Mol.include_substructures(substructures)
To exclude a given list of SMARTS strings substructures, run this command:
Mol.exclude_substructures(substructures)
After providing all requirements using the aforementioned functionalities, to generate molecules satisfying those requirements, run this command:
mols = Mol.solve(NumSolutions)
where NumSolutions is the number of generated molecules.
Shiqiang Zhang. Funded by an Imperial College Hans Rausing Scholarship and BASF SE Ludwigshafen am Rhein.
Christian Feldmann. Funded by BASF SE Ludwigshafen am Rhein.