Evaluating the generalizability of graph neural networks for predicting collision cross section

This repository contains code and data described in detail in our paper (Engler et al., 2024).

Table of Contents

• Citation
• Reproducibility
• Data

Citation

If you have found our manuscript useful in your work, please consider citing:

Engler Hart*, C., Preto*, A. J., Chanana*, S., Healey, D., Kind, T., Domingo-Fernandez, D. (2024). Evaluating the generalizability of graph neural networks for predicting collision cross section. Journal of Cheminformatics, 16 (105), 2024. https://doi.org/10.1186/s13321-024-00899-w

Reproducibility

Installation for development

poetry install
poetry run pre-commit install

Train the models

See the commands in the Makefile to train the models. Run them as make train-metlin-test-metlin

poetry run python scripts/train-test.py \
	--prefix "train-metlin-test-ccsbase" \
	--train-input-file "ccs-prediction/metlin_train_3d.parquet" \
	--test-input-file "ccs-prediction/ccsbase_3d.parquet" \
	--parameter-path "parameter/parameter-train-metlin-test-metlin.json" \
	--model-output-file "model/train-metlin-test-metlin.h5" \
	--coordinates-column-name "coordinates" \
	--coordinates-present \
	--smiles-column-name "smiles" \
	--adduct-column-name "adduct" \
	--ccs-column-name "ccs" \
	--dropout-rate 0.1 \
	--epochs 400 \
	> train-metlin-test-ccsbase.out 2>&1

• prefix is used to generate the output files of the predictions of the test set
• train-input-file is the training set (see notebooks/data_processing/2_data_splits.ipynb for details on the format)/
• test-input-file test set (see notebooks/data_processing/2_data_splits.ipynb for details on the format)
• parameter-path path to the file generated storing the parameters of the model
• parameter-path path to the file generated storing the parameters of the model
• model-output-file path to the model file
• coordinates-column-name column name of the 3d coordinates for each smiles
• coordinates-present if the coordinates are present (if not given, the model will use the smiles to generate the 3d coordinates)
• smiles-column-name column name of the smiles
• adduct-column-name column name of the adduct
• ccs-column-name column name of the ccs
• dropout-rate dropout rate of the model
• epochs number of epochs to train the model

Reproduce results

Run the notebooks located in the notebooks corresponding to each analysis. There are two folders:

• data_processing: notebooks to process the METLIN-CCS and CCSBase and make the data splits. Note that you have to download the two databases according to their licenses and modify the paths/names on the notebooks.
• reproduce_figures: the name of the notebooks indicates which notebook can reproduce which figures of the manuscript manuscript.
• exploring_predictions: notebooks to explore the predictions of the models in detail

Data

Predictions are available and can be directly downloaded from . The files should be unzipped and placed in the data directory.

The original datasets are available here:

• CCSBase: https://ccsbase.net/query
• Metlin: https://metlin.scripps.edu/ Each user should download the raw database (as excel/csv) and read them in the two notebooks for each database located at https://github.com/enveda/ccs-prediction/tree/main/notebooks/data_processing. Each notebook reads the csv/excel and formats it according to the input of Mol2CCS.

Predicting CCS

Train the model based on your own training dataset with [wrapper_train] and predict with wrapper_predict function.

The baseline original repositories are:

• SigmaCCS: https://github.com/zmzhang/SigmaCCS
• GraphCCS: https://github.com/tingxiecsu/GraphCCS

References

Ross, D. H., Cho, J. H., and Xu, L. (2020). Breaking down structural diversity for comprehensive prediction of ion-neutral collision cross sections. Analytical chemistry, 92(6), 4548-4557. https://doi.org/10.1021/acs.analchem.9b05772

Baker, E. S., Hoang, C., Uritboonthai, W., Heyman, H. M., Pratt, B., MacCoss, M., et al. (2023). METLIN-CCS: an ion mobility spectrometry collision cross section database. Nature methods, 20(12), 1836-1837. https://doi.org/10.1038/s41592-023-02078-5

Guo, R., Zhang, Y., Liao, Y., Yang, Q., Xie, T., Fan, X., et al. (2023). Highly accurate and large-scale collision cross sections prediction with graph neural networks. Communications Chemistry, 6(1), 139. https://doi.org/10.1038/s42004-023-00939-w

Xie, T., Yang, Q., Sun, J., Zhang, H., Wang, Y., and Lu, H. Large-Scale Prediction of Collision Cross-Section with Graph Convolutional Network for Compound Identification.