Skip to content

Latest commit

 

History

History
241 lines (128 loc) · 12.9 KB

README.md

File metadata and controls

241 lines (128 loc) · 12.9 KB
Raw

This is the Source Code and Data Repository of the HCASE Method

Cite Us

Zahoránszky-Kőhalmi, G., Wan, K.K. & Godfrey, A.G. Hilbert-curve assisted structure embedding method. J Cheminform 16, 87 (2024). DOI: 10.1186/s13321-024-00850-z

Preprint

G. Zahoranszky-Kohalmi, K. K. Wan, A. G. Godfrey, Hilbert-Curve Assisted Structure Embedding Method, ChemRxiv, 28 February 2020, Version 1, DOI: 10.26434/chemrxiv.11911296.v1


HCASE Method

Publication Codebase and Continuous Development

The publication branch of the HCASE repository contains the source code and data files in the exact same status at the time the manuscript was under review and got accepted. We will keep this branch unchanged, but we continue the development and improvement of the codebase in the main branch.

If you are interested in contributing to the HCASE repository please refer to the For Contributors section below.

Test Drive HCASE

We prepared a workflow that showcases how you can create an HCASE space and embed compounds into it, potentially using multiple cores (recommended). It is available as examples/HCASE_Embedding_Parallel.ipynb in the root of the HCASE repository. The workflows does not contain the visualization step, for that we refer you to examples/publication_workflows/HCASE_Experiments_Cherry_Picked_Scaffolds.ipynb and examples/publication_workflows/HCASE_Experiments_KNN_in_HCASE_Space.ipynb as starting points.

The examples/HCASE_Embedding_Parallel.ipynb can be executed once completing the Prerequisites through Setting Up Conda Environment sections (see: below).

For Contributors

We appreciate your contribution to the source code of the HCASE method!

If you would like to create a new feature, or submit a fix, please follow this process. First, create a personal feature branch (e.g. feat_new_1) from the feature branch. Once your code is complete and tested OK on your end, please create a pull-request (PR) from your feature branch to the feature branch of the HCASE repository. Approved features will be merged to the main branch.

Process to Replicate the Analysis Enclosed in the HCASE Manuscript

Prerequisites

Make sure that Git is installed along with Git Large File Support (LFS)

To install GIT LFS use:

git lfs install --skip-smudge

Conda environment manager for Python is required

See: https://conda.io/projects/conda/en/latest/user-guide/install/index.html

Setting Up Repository

Clone Repository

git clone https://github.com/ncats/hcase

Pull Large Files

cd hcase

git lfs pull (this may take some time)

Repository Structure

The primary HCASE library code is within the hcase subdirectory. See the pip installion section for imformation on how to install this code via pip.

Examples of how to use this code are found within the examples subdirectory. This directory contains a conda environment which will install the hcase code automatically.

Installing as a library

The latest HCASE code can be installed via pip with the following command:

pip install git+https://github.com/ncats/hcase.git

Additionally certain branches can be installed with:

pip install git+https://github.com/ncats/hcase.git@BRANCH-NAME

Setting Up Examples Conda Environment

Run the following to create the Conda environment:

conda env create -f examples/environment.yml (this step needs to be done only once, if you don't already have the hcase_examples conda environment)

conda activate hcase_examples (this step is necessary everytime you want to run the HCASE method and/or the experiments)

Post Publication Notes

For the sake of clarity, we moved all the Jupyter Notebook (.ipynb) files into examples/publication_workflows subdirectory after our paper got published.

To view the repository code at the time of publication view the publication branch.

Executing Individual Steps

Alternatively each seperate experiment can be ran by running the following jupyter notebooks. All jupyter notebooks are within the examples or examples/publication_workflows directories.

  • Step 1: Create HCASE space using ChEMBL scaffolds

Run workflow in HCASE_Embedding.ipynb .

  • Step 2: Embed approved drug molecules from DrugBank into HCASE ChEMBL Scaffold space, and visualize the kNNs (k=5) of select chemotypes. Additionally, this is compared to the position of the same select chemotypes and 25 randomly selected molecules.

Run workflow in HCASE_Experiments_KNN_in_HCASE_Space.ipynb .

  • Step 3: HCASE Embedding of Cherry-Picked Scaffolds and their 50+50 Immediate Neighbors Based on Scaffold-Key Ordering

Run workflow in HCASE_Experiments_Cherry_Picked_Scaffolds.ipynb .

  • Step 4: Create the embedding of CANVASS molecules into HCASE ChEMBL space.

Run workflow in HCASE_Experiments_CANVASS_into_HCASE_ChEMBL.ipynb .

  • Step 5: Create the embedding of approved drug and CANVASS molecules into HCASE NatProd space.

Run workflow in HCASE_Experiments_Drugs_and_CANVASS_into_HCASE_NatProd.ipynb .

  • Step 6: Comparing HCASE ChEMBL and NatProd embeddings of approved drugs and CANVASS compounds

Run workflow in HCASE_Experiments_Compare_HCASE_ChEMBL_NatProd_Embeddings.ipynb .

  • Step 7: Quantify overlap of approved drugs and CANVASS compounds embeddings in HCASE ChEMBL and NatProd spaces

Run workflow in HCASE_Experiments_Quantify_Overlap.ipynb .

  • Step 8: Chebyshev Distance and Rank Distance Correlation.

Run workflow in HCASE_Experiments_Chebyshev.ipynb .

  • Step 9: T-SNE Embedding of drugs and highlighting KNNs (k=5) of select molecules. Repeating the embedding and visualization on a reduced set (90%) of drugs.

Run workflow in HCASE_Experiments_tSNE.ipynb .

  • Step 10: Creating an HCASE space using a reduced set (90%) of ChEMBL scaffolds, and the previously cherry-picked scaffolds.

Run workflow HCASE_Embedding_Reduced.ipynb .

  • Step 11: Creating the HCASE embedding of the cherry-picked scaffolds into the reduced HCASE ChEMBL space generated in the previous step.

Run workflow HCASE_Experiments_Cherry_Picked_Scaffolds_in_Reduced_HCASE_Space.ipynb .

  • Step 12: Repeating Step 2, only the KNN analysis of approved drug molecules is performed in a reduced HCASE ChEMBL scaffold space (90% subset of original ChEMBL scaffold used to create the HCASE space). So-called Not-NN neighbor analysis is not repeated, only the KNNs of the randomly selected 5 drug molecules.

Run workflow HCASE_Experiments_KNN_in_Reduced_HCASE_Space.ipynb .

  • Step 13: Perform embedding of cherry-picked scaffolds using the Altered t-SNE method into ChEMBL scaffold space.

Run workflow HCASE_Experiments_Altered_tSNE_CherryPicked_Scaffolds.ipynb .

  • Step 14: Repeating the previous step (Step 13) utilizing the union of the 90% subset of the reference scaffolds and the Cherry-Picked scaffolds.

Run workflow HCASE_Experiments_Altered_tSNE_Reduced_HCASE_Space_CherryPicked_Scaffolds.ipynb .

  • Step 15: Embed approved drug molecules from DrugBank into Altered t-SNE ChEMBL Scaffold space, and visualize the kNNs (k=5) of select chemotypes.

Run workflow HCASE_Experiments_Altered_tSNE_KNN.ipynb .

  • Step 16: Repeating Step 15, i.e. performing the Altered t-SNE embedding of the same 5 randomly selected drug molecules and their KNNs with the difference that a 90% reduced Altered t-SNE scaffold space is used for the embedding.

Run workflow HCASE_Experiments_Reduced_Altered_tSNE_KNN.ipynb .

  • Step 17: Aggregate embeddings.

Run workflow HCASE_Experiments_Aggregate.ipynb .

License Related Section

This repository contains source code, Jupyter notebooks, data and results files which are organized into various subdirectories.

  • Source code subdirectories:

hcase, code/, and aux_code

  • Data and results subdirectory:

data/

  • Jupyter Notebooks:

Jupyter Notebooks can be found in the code/, examples, and data/execute_notebooks/ subdirectories (files of .ipynb extension).

  • Workflows:

KNIME workflows can be found in workflows/ subdirectory.

  • Source Code License of HCASE Repository

The applicable license to source code can be found under filename: code/LICENSE and aux_code/LICENSE (same license types, i.e. [https://opensource.org/licenses/MIT](MIT License)). This license is applicable to all files recursively in the source code subdirectories as defined above. The files code/NOTES and aux_code/NOTES list source code modules that were utilized and their respective licenses. These modules have their own licenses which might be different from the Source Code License of this repository, and they need to be respected accordingly.

  • Data License of HCASE Repository

The applicable license to data and results can be found under: data/LICENSE and plots/LICENSE ([https://creativecommons.org/licenses/by/4.0/legalcode.txt](Creative Commons Attribution 4.0 International Public License CC-BY 4.0 International)). This license is applicable to all files recursively in the data and results subdirectory as defined above. The files listed in data/NOTES and plots/NOTES lists input files and resources utilized to perform the experiments and are considered as derivative work of those resources. These input files and resources have their own licenses which might be different from the Data License of this repository, and they need to be respected accordingly. In the same file we also list which results files can be considered as derivative works, and we also list the the respective ascendent data source(s).

  • Jupyter Notebook License of HCASE Repository

Jupyter Notebboks are special in the sense that they are comprised of source code, but they can also contain embedded data and plot (graph) sections. This duality is resolved via dual-licensing as follows. The code sections of Jupter Notebooks fall under the same license as source codes, i.e. the MIT License code/LICENSEwhereas data and plot sections embedded into the Jupyter Notebooks fall under the same license as data and result files, i.e. Creative Commons Attribution 4.0 International Public License CC-BY 4.0 International License data/LICENSE. Remarks enclosed in the code/NOTES file are also valid for code section of the Jupyter Notebooks. Remarks enclosed in the data/NOTES and plots/NOTES files are also valid for the embedded data and plots of the Jupyter Notebook files.

  • Workflow License of HCASE Repository

KNIME workflows utilized in this study are contained in the workflows/ directory, and are subject of the "KNIME License Terms and Conditions" see: workflows/LICENSE.

Remarks

Exporting Conda Environment for Reproduction

Critical dependencies :

  • hilbertcurve
  • networkx
  • rdkit

References


https://stackoverflow.com/questions/5226311/installing-specific-package-versions-with-pip?rq=1
https://stackoverflow.com/questions/38411942/anaconda-conda-install-a-specific-package-version
https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html
https://stackoverflow.com/questions/41274007/anaconda-export-environment-file
https://stackoverflow.com/questions/43222407/how-to-list-package-versions-available-with-conda
https://www.markdownguide.org/basic-syntax/
https://stackoverflow.com/questions/62852772/error-when-getting-git-lfs-files-object-does-not-exist-on-the-server-404-obj