DeepRMethylSite: Prediction of Arginine Methylation in Proteins using Deep Learning

DeepRMethylSite is a deep-learning based method for Arginine Methylation sites prediction in proteins. It is implemented using Keras (version 2.2.4) and Tensorflow (version 1.15) backend and has been tested on both in Windows and Linux OS.

Pre-requisites

Python 3.6
Tensorflow (>version 1.15) - (Optional: Tensorflow GPU for GPU based machine)
Keras (>version 2.2.4) - (Optional: Keras GPU for GPU based machine)
Numpy
Biopython
Sklearn
Imblearn
Winrar or any compression program to open .rar file

Running on CPU or GPU

To run in CPU, installation of Tensorflow and Keras will suffice. However, to run in GPU, further Tensorflow-gpu and keras-gpu must be installed. Tensorflow GPU and Keras GPU version utilizes cuda cores in our GPU (in our case NVIDIA 2080 TI) for faster training time. However, running in GPU is not mandatory.

Dataset

Dataset is in fasta format. Both training and testing datasets are provided which are independent (one does not include others). Training dataset for positive and negative are train_s33_Pos_51.fasta and train_s33_Neg_51.fasta respectively. Testing dataset for positive and negative are test_s33_Pos_51.fasta and test_s33_Neg_51.fasta respectively. Training dataset is made available so that future models can be trained for the comparison purpose.

Model

The best trained model for both CNN and LSTM (used in our final results) have also been included. The model_best_cnn.h5 is the best trained model for CNN and model_best_lstm.h5 is the best trained model for LSTM respectively.

Code

Grid search ensemble code for testing is provided i.e. model_gridsearch_load.py. It requires following files to run: -model_best_cnn.h5 -model_best_lstm.h5 -test_s33_Pos_51.fasta -test_s33_Neg_51.fasta The output of this is the result mentioned in our research paper.

Prediction for given test dataset (Procedure)

  - Download all rar files from model.part01 to model.part07 and keep in the same folder.
  - Open model.part01 with compression tool like winrar and extract it. This will extract both model files from our 
    research, model_best_cnn.h5 and model_best_lstm.h5.
  - Download test datasets, test_s33_Pos_51.fasta and test_s33_Neg_51.fasta, and python code model_gridsearch_load.py.
    Keep them in the same folder as model files.
  - Run model_gridsearch_load.py and you will get output mentioned in our research.
  - In linux code will be, $python3 model_gridsearch_load.py

Prediction for your dataset

If you would like to use DeepRmethylSite to predict Arginine Methylation sites in the protein of your interest, you should prepare your dataset in the same format as the test dataset which is basically a FASTA format. This model works for window size 51 only, meaning for the residue of your interest you should provide 25 resiudes downstream and 25 residues upstream. e.g. if you want to predict whether the 'Arginine' residue in Position 735 in protein Q4KWH8 is methylated or not, the input file should contain 25 residues upstream of R (position 735 in protein Q4KWH8) and 25 residues downstream of R.

The general format for your dataset should be:

sp|Q4KWH8|PLCH1_HUMAN%730%755
PKKQLILKVISGQQLPKPPDSMFGDRGEIIDPFVEVEIIGLPVDCCKDQTR

Citation:

Please cite the following paper if you use DeepRMethylsite. DeepRMethylSite: Meenal Chaudhari*, Niraj Thapa*, Kaushik Roy, Robert H. Newman, Hiroto Saigo, Dukka B. KC, DeepRMethylSite: Prediction of Arginine Methylation in Proteins using Deep Learning, Mol. Omics, 2020,16, 448-454.

Contact

Feel free to contact us if you need any help : [email protected]