RNAgranuleModel

Package installation (Python 3.8.8)

Biopython package (Version: 1.80); Scipy package (Version 1.9.3); scikit-learn package (Version 1.3.0); dcor package (Version 0.6)

Model training

This section provides step-by-step instructions to train the RNA granule prediction models.

Prerequisites

1. Download the dataset 'RNAgranule_model_training_data.zip' from the ResearchGate (https://www.researchgate.net/publication/386099818_Training_data_for_the_RNA_granule_model).
2. Clone or download all files from the 'model training' directory in this repository.

Setup

3. Extract the dataset and place all files (from both the dataset and repository) in a single directory. Your directory structure should look like: project_directory/
   ├── aa_features_pre.py
   ├── data_processing.py
   ├── fasta_transfer.py
   ├── IDR_identify.py
   ├── model_train.py
   ├── my_model_SG_1.pickle - my_model_SG_10.pickle
   ├── pdb30.csv
   ├── physicochemical_property_cal.py
   ├── RNA_granule_2Kmer_tier1_result.csv
   ├── RNA_granule_3Kmer_tier1_result.csv
   ├── RNA_granule_data.xlsx
   ├── SG_auc_df.csv
   ├── total_aa.csv
   ├── total_data.csv
   ├── total_data.fasta
   ├── uniprot_human_proteome.csv

Training Models

4. Run the model training script with the appropriate parameters:
   For Stress granule model: Set target = 'SG' in Line 189 and tier = 1 in Line 190
   For P-body/Stress granule model: Set target = 'PBSG' in Line 189 and tier = 1 in Line 190
   For P-body model only: Set target = 'PB' in Line 189 and tier = 2 in Line 190
   Execute the training by running:
   python model_train.py

Predicting

The dataset for predicting can be downloaded in https://www.researchgate.net/publication/386100715_RNAgranule_model_predicting_data. The codes in the 'Predicting' are used for predicting new proteins.

1. To predict the human proteome You can run the 'model_predict_proteome.py' file.
2. To predict your own proteins You can replace the information of each column in the 'uniprot_human_proteome.csv' file.

Functional dense PPI cluster analysis

You can find the whole RNA granule proteome PPI community in the SI.xlsx file. To visualize the identified functional dense PPI clusters:

1. you can download the SI.xlsx file
2. collect the cluster 1, cluster 2 and cluster 3 protein (prob>0.5) lists, OR cluster 1-, cluster 2- and cluster 3- protein (prob≥0.7) lists in each sheet
3. input the protein names of the 'N_1' column in the STRING website (Version: 12.0, link: https://version-12-0.string-db.org/).
4. use default basic settings to visualize and analyze your proteins of interest and see the interactions.

---

Feel free to send me an email if you need any help about the codes. The files are under the CC BY-NC license.