tawssie / zmpy3d_pt

Python implementation of 3D Zernike moments with PyTorch

ZMPY3D_PT

Update: ZMPY3D PyTorch implementation is available (August 25, 2024).

• 3D Zernike moments with PyTorch:
• Shape similarity with PyTorch:
• Structure superposition with PyTorch:

ZMPY3D: accelerating protein structure volume analysis through vectorized 3D Zernike Moments and Python-based GPU Integration

For CPU support only, please refer to the repository:

ZMPY3D supports NumPy (https://github.com/tawssie/ZMPY3D)

For GPU support with TensorFlow, CuPy and PyTorch, please refer to the other three repositories:

ZMPY3D_TF supports Tensorflow (https://github.com/tawssie/ZMPY3D_TF)

ZMPY3D_CP supports CuPy (https://github.com/tawssie/ZMPY3D_CP)

ZMPY3D_PT supports PyTorch (https://github.com/tawssie/ZMPY3D_PT)

Here presents a Python-based software package, ZMPY3D, to accelerate the moments computation by vectorizing the mathematical formulae, enabling their computation in graphical processing units (GPUs). The package offers popular GPU-supported libraries such as CuPy and TensorFlow along with NumPy implementations, aiming to improve computational efficiency, adaptability, and flexibility in future algorithmic development.

Installation

Prerequisites:

• ZMPY3D : Python >=3.9.16, NumPy >=1.23.5
• ZMPY3D_CP: Python >=3.9.16, NumPy, CuPy >=12.2.0
• ZMPY3D_TF: Python >=3.9.16, NumPy >=1.23.5, Tensorflow >=2.12.0, Tensorflow-Probability >=0.20.1
• ZMPY3D_PT: Python >=3.9.16, NumPy >=1.23.5, PyTorch >= 2.3.1

1. Open the terminal
2. Using pip to install the package through PyPI
3. Run pip install ZMPY3D_PT for the installation

Usage

• 3D Zernike moments with Tensorflow:
• Shape similarity with CuPy:
• Structure superposition with NumPy:
• Runtime evaluation:

Performances

A voxel cube with dimensions of 100x100x100 was applied to perform 10,000 3D Zernike moment calculations, using 2 different maximum orders 20 and 40. Execution times for different hardware configurations using TensorFlow, CuPy, and NumPy libraries:

NumPy

Order	CPU1	CPU2
20	33m20s	14m1s
40	951m40s	338m20s

TensorFlow

Order	T4	RX3070Ti	V100	L4
20	1m1s	0m36s	0m31s	0m39s
40	24m40s	9m3s	10m54s	11m13s

CuPy

Order	T4	RX3070Ti	V100	L4
20	4m45s	2m30s	1m42s	2m50s
40	35m20s	19m19s	14m45s	18m40s

Note: m = minutes, s = seconds.

Cache data for order 40

Due to GitHub's file size limitations, follow these steps to download the cache data for order 40 (1.3G) in the ZMPY3D_PT package:

1. Locate Package Folder

• Open your terminal and execute the following command to find the folder of the ZMPY3D_PT package:
• python -c "import ZMPY3D_PT; print(ZMPY3D_PT.__file__)"
• Note the path, which ends with /User/path/ptyhon/site-packages/ZMPY3D_PT/__init__.py.

2. Navigate to Cache Data Folder

• Go to the cache_data folder at the same level as __init__.py file, i.e., /User/path/ptyhon/site-packages/ZMPY3D_PT/cache_data.

3. Download the Cache File:

• Download the 1.3 GB max order 40 .pkl file to the cache_data folder from the link below. https://drive.google.com/uc?id=1RR1rF_5YJqaxNC5AK0Ie_8MswGb0Tttw

Further reading: What can 3D Zernike moments do?

• Enhancing fold classification
  • Real-time structure search and structure classification for AlphaFold protein models
  • Real time structural search of the Protein Data Bank
• Facilitating structural superpositions
  • ZEAL: Protein structure alignment based on shape similarity
• Supporting protein docking
  • Protein-protein docking using region-based 3D Zernike descriptors
• Assisting molecular dynamics
  • Binding site identification of G protein-coupled receptors through a 3D Zernike polynomials-based method: application to C. elegans olfactory receptors
  • Quantitative characterization of binding pockets and binding complementarity by means of zernike descriptors
• Enabling structure-based virtual screening
  • PL-PatchSurfer3: Improved Structure-Based Virtual Screening for Structure Variation Using 3D Zernike Descriptors
• Forecasting interacting interfaces
  • Antibody interface prediction with 3D Zernike descriptors and SVM
  • Exploring the potential of 3D Zernike descriptors and SVM for protein-protein interface prediction

Contributing

Feel free to submit pull requests for improvements or bug fixes.

---

Citation

Lai, J. S., Burley, S. K., & Duarte, J. M. (2024). ZMPY3D: Accelerating protein structure volume analysis through vectorized 3D Zernike moments and Python-based GPU integration. (Bioinformatics Advances, vbae111, https://doi.org/10.1093/bioadv/vbae111)

License

This project is licensed under the GNU General Public License v3.0. You can view the full license here.