SCC plotting

(c) Jonathan Goodman, Richard Lewis, Benji Rowlands
2024
jmg11@cam.ac.uk
University of Cambridge

Overview

This repository contains scripts to reproduce the SCC plots from the paper "Towards automatically verifying chemical structures: the powerful combination of ¹H NMR and IR spectroscopy".

The main plotting script is contained in scc.py. An example input file is contained in config.yaml.

The softmax scaling script is contained in scale_scores_softmax.py.

Requirements

• Operating system: this software has been tested on macOS Sonoma 14.5. It should work on other operating systems, but has not been tested on them.
• Python version: this software was developed using Python 3.12.

Installation

To run this software, ensure you have Python 3.12 installed along with the following packages:

• matplotlib (version 3.9.0)
• pandas (version 2.2.2)
• numpy (version 2.0.1)
• scipy (version 1.14)
• scikit-learn (version 1.5.1)
• scienceplots (version 2.1.1)
• pyyaml (version 6.0.2)
• PyQt5 (version 5.15.11)

Installation time will depend on how long it takes to download the necessary packages, but should be less than a minute.

To configure the environment correctly, follow the following steps:

1. Install Python: Ensure you have Python 3.12 installed. You can download it from the official Python website.

2. Create a virtual environment:

   python3.12 -m venv .venv
   source .venv/bin/activate 
   

3. Activate the virtual environment and install the required package:

   source .venv/bin/activate
   pip install matplotlib==3.9.0 pandas==2.2.2 numpy==2.0.1 scipy==1.14 scikit-learn==1.5.1 scienceplots==2.1.1 pyyaml==6.0.2 pyqt5==5.15.11
   

Usage

SCC plots

1. Prepare your input file: The input file contains the settings for the plot, and should be named config.yaml. An example input file is provided with all the available settings shown. The path to the data needs to be specified in the input file.

2. Run the script: Execute the script from the command line. With no modification to the input file provided, a plot will be generated using data from the demo/ directory.

python scc.py

Execution time should be a few seconds.

Softmax scaling

scale_scores_softmax.py takes 4 required command line arguments:

• --target: Path to the scaling target, i.e. the method whose standard deviation is to be matched
• --target_output: Softmax will be applied to the scores of the scaling target, and the new scores saved to this path
• --input_file: Path to the data to be scaled
• --output_file: Save path for the scaled data

The scaling can be demonstrated using data from the demo/ directory:

cd demo
python ../scale_scores_softmax.py --target IR.Cai.csv --target_output IR.Cai_softmax.csv --input_file DP4.csv --output_file DP4_softmax.csv

Execution time should be at most a few seconds.

Output

scc.py will save a pdf of the SCC plot if save_name and save_dir are both specified in config.yaml. Otherwise, the plot will simply be shown.

scale_scores_softmax.py will save csv files containing softmax-scaled data to the paths specified by the --target_output and --output_file arguments.

Example data

demo contains 3 files:

• DP4.csv contains DP4* scores for the molecules in the test set
• IR.Cai.csv contains IR.Cai scores, obtained using high-level theory and a hwhm of 12cm^-1
• ACD.csv contains ACD scores for the molecules in the test set

The data are in csv format. The 'Molecule' column identifies the molecule in question from the test set. The 'Comparison' column identifies which isomers are being compared. The '0', '1', '2', and '3' columns contain the scores for the relevant isomer for each molecule. As throughout this work, the score in the '0' column is the score for the correct isomer; the other columns contain scores for incorrect isomers.

Contact

For any questions or issues, please contact Jonathan Goodman at jmg11@cam.ac.uk.