Quantitative Structure-Property Relationships for Liquid Crystals
This is a companion code for the paper
- Machine Learning Analysis for the Contribution of Molecular Fine Structures to the Symmetry of Liquid Crystals, by Yoshiaki Uchida, Shizuo Kaji, Naoto Nakano
Requirements
First, to set up Python environment, install Anaconda.
Install necessary libraries. The actual command for installation depends on the environment. For example,
> pip install iterative-stratification rdkit mordred pyarrow
> pip install lightgbm
> conda install pytorch torchvision torchaudio torcheval pytorch-cuda=11.8 -c pytorch -c nvidia
> conda install pyg -c pyg
> pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.0.0+cu118.html
How to use
-
Prepare a csv file containing at least the following two columns
- ID: integer index to identify molecules uniquely
- SMILES: SMILES one-line notation
- Phases: Phase sequence in the format of LiqCryst
a small sample file is included as sample.csv.
-
Open the Jupyter notebook LC_QSPR.ipynb and follow the instructions.
-
First, the database containing descriptors and phase transition temperatures should be created from the csv file above. Look at the Compute descriptors from SMILES and parse Phase sequence section.
-
There are two main machine learning models: GBM and GNN. They can be tried separately.