Skip to content

Latest commit

 

History

History
79 lines (54 loc) · 3.97 KB

README.md

File metadata and controls

79 lines (54 loc) · 3.97 KB
Raw

BCrystal

BCrystal: An Interpretable Sequence-Based Protein Crystallization Predictor

Motivation

Protein crystallization allows to study molecular structure. Novel in silico, accurate, sequence-based protein crystallization predictors are highly sought.

Installation

Requirements

This step will install all the dependencies required for running BCrystal. You do not need sudo permissions for this step.

  • Install Anaconda

    1. Download Anaconda (64 bit) installer python3.x for linux : https://www.anaconda.com/distribution/#download-section
    2. Run the installer : bash Anaconda3-2019.03-Linux-x86_64.sh and follow the instructions to install.
    3. Install xgboost: conda install -c conda-forge xgboost
    4. Install shap: conda install -c conda-forge shap
    5. Install Bio: conda install -c anaconda biopython

  • R requirements

    • Run R REPL by running the following: R
    • Install R libraries
      1. Interpol (do install.packages('Interpol') )
      2. bio3d (do install.packages('bio3d') )
      3. doParallel (do install.packages('doParallel'))
      4. zoo (do install.packages('zoo'))

    Quit R REPL: quit()

  • SCRATCH (version SCRATCH-1D release 1.2) (https://scratch.proteomics.ics.uci.edu, Downloads: https://download.igb.uci.edu/#sspro)

    1. Run wget https://download.igb.uci.edu/SCRATCH-1D_1.2.tar.gz
    2. Run tar -xvzf SCRATCH-1D_1.2.tar.gz
    3. Run cd SCRATCH-1D_1.2
    4. Run perl install.pl
    5. Run cd ..
    6. Replace the blast in SCRATCH-1D_1.2/pkg/blast-2.2.26 with a 64 bit version of blast-2.2.26 if you are running on a 64 bit machine (ftp:https://ftp.ncbi.nlm.nih.gov/blast/executables/legacy.NOTSUPPORTED/2.2.26/).

  • DISOPRED (version 3.16) (https://bioinfadmin.cs.ucl.ac.uk/downloads/DISOPRED/)

    1. Run wget https://bioinfadmin.cs.ucl.ac.uk/downloads/DISOPRED/DISOPRED3.16.tar.gz
    2. Run tar -xvzf DISOPRED3.16.tar.gz
    3. Run cd DISOPRED/src/
    4. Run make clean; make; make install
    5. In run_disopred.pl file within the DISOPRED folder put my $NCBI_DIR = <path-to-SCRATCH-folder>/pkg/blast-2.2.26/bin
    6. In run_disopred.pl file also put my $SEQ_DB = <path-to-SCRATCH-folder>/pkg/PROFILpro_1.2/data/uniref50/uniref50.

DISOPRED and SCRATCH-1D_1.2 should be in the same directory as Data folder. Data folder has the training and the 3 test set proteins in fasta format as well as files corresponding to their true labels - crystallized (1) or not (0).

Run BCrystal on New Test file

To run BCrystal on your own protein sequences you need the following three things:

  1. Protein Sequence File: Protein sequence/sequences of interest in fasta format (https://en.wikipedia.org/wiki/FASTA_format). We provide Data/test.fasta as sample test file.
  2. SCRATCH: Software used to extract structural features from a given protein sequence file. Follow instructions in the previous section to install SCRATCH.
  3. DISOPRED: Software used to extract disorder features from a given protein sequence file. Follow instructions in the previosu section to install DISOPRED.

Execute in the command line

  1. Rscript --vanilla features_PaRSnIP_v2.R <your-test>.fasta
  2. python xgb.py features.csv <your-test>.fasta <output_folder>

In the <output_folder> you will find 2 outputs:

  1. prediction.csv - Containing the crystallization propensity
  2. bar_plot_i.png - where i=1 if a solo sequence is passed in fasta otherwise its the nth sequence in the test fasta file.

Run BCrystal in Train Mode

To run BCrystal for training xgboost model on our training proteins, you need to do the following:

Execute in the command line

  1. Rscript --vanilla features_PaRSnIP_v2.R Data/Train/FULL_Train.fasta
  2. python xgb_train.py

The training set is readily available for ease of use at: https://drive.google.com/file/d/1FRWIcs4xvK2O5OCqhg7u5g_4qm2zJn2d/view and can be used in combination with Step 2 to generate the BCrystal model.

Your output will be a file called train.model