KANPM-DTA: Improving Drug-Target Affinity Prediction with Kolmogorov-Arnold Networks and Pre-trained Models
Figure 1. The proposed KANPM-DTA model architecture. (A) Graph and sequence-based protein feature encoding. (B) Graph and sequence-based drug feature encoding. (C) Adaptive feature learning between drug and protein graph-sequence features.
- Python==3.9.21
- numpy==2.0.2
- pandas==2.2.3
- torch==2.6.0
- transformers==4.49.0
- rdkit==2024.3.2
- fair-esm==2.0.0
git clone https://github.com/khanonuvov/KANPM-DTA.git
cd KANPM-DTA
For each dataset (davis, kiba, metz, bindingdb, Lung Cancer Test) run these three commands or download the pretrained models from here.
python pretrained/chemberta_pretraiend.pypython pretrained/esmC_pretraiend.pypython pretrained/esm2_map.pypython code/cold_split.pypython code/train.pypython code/pred.pyIf you find our paper or code useful in your research, please consider citing:
@article{10.1093/bib/bbag112,
author = {Rakib, MD Youshuf Khan and Alamin, Muhammad Habibulla and Li, Jiamu and Mamun, Sheikh Sohan and Gobena, Kaleb Amsalu and Ren, Shengbing},
title = {KANPM-DTA: improving drug–target affinity prediction with Kolmogorov–Arnold networks and pretrained models},
journal = {Briefings in Bioinformatics},
volume = {27},
number = {2},
pages = {bbag112},
year = {2026},
month = {03}
}For inquiries, please contact MD Youshuf Khan Rakib (Email: khanushuf4619@csu.edu.cn).