PMAW

Source code of the paper "Prediction of Molecular Absorption Wavelength Using Deep Neural Networks"

Jinning Shao¹, Yue Liu¹, Jiaqi Yan¹, Ze-Yi Yan, Yangyang Wu, Zhongying Ru, Jia-Yu Liao*, Xiaoye Miao*, Linghui Qian*

Installation

For GCN and AttentiveFP,

conda install --yes --file deepchem_requirements.txt

For other models (FCNN, CNN, Chemfluo),

conda install --yes --file molfp_requirements.txt

Prepare data

Make sure your data format belongs to {descriptor, morgan, maccs, smiles and morgan+maccs}. In all formats, the first column should denote the true absorption wavelengths. For more details, please refer to our data.

Inference

We didn't prepare the model weights here, since the model is lightweight and easy to train. You can now put arbitrary values in the first column of data. Suppose you have put the morgan data in /src/data/database. First, you should modify the data path here; Then run

python /src/mlp_morgan_out.py 

The inference result will be put in /src/Out/Large_MRE_out_points.csv

Tips: You may use mlp_descriptor_out.py for descriptor datasets, and et cetera.

Train your own data

Suppose you hace put the morgan training data in /src/data/database. You should modify the data path here. Besides, you may add your codes to save the model weights here.