proteinpicker's Introduction

ProteinPicker

Dependencies

The synthetic tomograms (represented as MRC files) used for our machine learning models have been generated using cryo-TomoSim. The software can be downloaded at their Github repository at: https://github.com/carsonpurnell/cryotomosim_CTS

PDB Files

There are three protein classes that our protein classification model has been trained on. Their AlphaFold entries were provided in the paper, but their entries can also be found here for convenience:

Protein KIAA0100 (BLTP2): https://alphafold.ebi.ac.uk/entry/Q14667
Coagulation Factor V: https://alphafold.ebi.ac.uk/entry/P12259
Telomere-associated protein RIF1: https://alphafold.ebi.ac.uk/entry/Q5UIP0

Structure

This Github repository holds all of the code used to train the atlas and tomogram models featured in our CS562 Final Project. As a general overview of how the repository is structured:

archive: holds various code files that are no longer used by the ProteinPicker project
imgs: holds 2D slices of MRC files, back from when the model was originally a 2D CNN model. These images are no longer used for the updated 3D CNN model.
mrc: holds all of the mrc files used for the 3D CNN model
- atlas-mrc holds all of the atlas MRC files used for training the atlas 3D CNN model
- tomogram-mrc holds all of the inverted MRC files used for training the inverted 3D CNN model
saved_model: stores all models featured in the ProteinPicker project. For the purposes of the final project, there are only two models that were featured in the paper.
- atlas_3d
  - atlas_3d_empty_model.keras: "atlas" model featured in the paper
  - Atlas3DModel.ipynb: Jupyter notebook used to create the aforementioned model
- tomogram_3d:
  - tomogram_3d_empty_model.keras: "inverted" model featured in the paper
  - Tomogram3DModel.ipynb: Jupyter notebook used to create the aforementioned model

Recommend Projects