MapLE: Matching Molecular Analogues Promptly with Low Computational Resources by Multi-Metrics Evaluation

This general strategy aims to promptly match analogous molecules with low computational resources by multi-metrics evaluation. If you want to know the main ideas and results of the work, please read this student abstract published on AAAI-2024.

Usage

Getting Raw Data

Please go to the official website of PDBbind dataset to download raw data(Welcome to PDBbind-CN database), the CASF-2016 version is used in this work. Then place the downloaded data in the project's"mol_data/pre_process" directory.

Due to the large size of the original data set, this project provides some samples as a demo in the project's "mol_data/sample" directory.

Preprocess

Use the pre_process.py file to process the raw data.

After processing, we can get multi-metrics inverted lists which are used in the next step.

Progressive prompt evaluation

After setting the dataset directory, please directly use the match.py file to match analogous molecules.

Reference:

If you find our work useful, please consider citing:

@Article{Chen2024,
  author  = {Chen, Xiaojian and Liao, Chuyue and Gu, Yanhui and Li, Yafie and Wang, Jinlan and Chen, Yi and Kitsuregawa Masaru},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  title   = {MapLE: Matching Molecular Analogues Promptly with Low Computational Resources by Multi-Metrics Evaluation (Student Abstract)},
  year    = {2024},
  volume  = {38},
  url     = {https://ojs.aaai.org/index.php/AAAI/article/view/30427},
  DOI     = {10.1609/aaai.v38i21.30427},
  number  = {21},
  pages   = {23456-23457}
}