DeepHelicon

tags: transmembrane protein residue contact predictor inter-helical contacts v1.0

Overview

DeepHelicon is a predictor for accurately predicting inter-helical residue contacts in transmembrane proteins. This repository provides a standalone package of DeepHelicon.

System Requirement

The software is only allowed to be run on a Linux operation system. Please be sure of python (version>3.5) installed before using. We highly recommend a package Anaconda, an integrated development environment of python, which eases the use and management of python packages.

Installation

1. install the dependencies and specify where their executables are.

   • HHblits - generating multiple sequence alignments
   • CCMpred - predictor of residue contacts
   • Gaussian DCA - predictor of residue contacts
   • Freecontact - predictor of residue contacts
   • plmDCA - predictor of residue contacts
   • TMHMM2.0 - predictor of transmembrane segments
   • EVcouplings - python interface used for predicting protein structure, function and mutations using evolutionary sequence covariation
   • Uniprot database - a curated protein sequence database for hhblits

   NOTE

   1. Download either the recommended Uniprot database, or other database curated for HHblits and make it work with hhblits.
   2. Download and specify the path of EVcouplings package in src/Plmc_dhc_rs.py. Please do not rename this package.
   3. Please add the paths of the above executable programs to src/troll.sh.

2. install DeepHelicon

   • To download a released package of DeepHelicon stable version, click here.

   • To download all prediction models of DeepHelicon click here.

   • To obtain the latest version of DeepHelicon, do

   git clone https://github.com/2003100127/deephelicon.git
   

3. install python dependencies

   pip install -r requirements.txt
   

Usage

1. part 1 of feature generation via src/troll.sh

   • description troll.sh is used to generate transmembrane topologies and most of evolutionary coupling features, including CCMpred, EVfold, plmDCA.

   • name format of a FASTA file Note, the FASTA file you are providing should have the suffix '.fasta'. For example, a whole FASTA file name can be '2wsc2.fasta' or a CASP name 'T1024.fasta'.

   • shell commands

     • general

     ./troll.sh -n NAME -c CHAIN -i /INPUT/PATH/
     

     • example

     ./troll.sh -n 2wsc -c 2 -i ./input/
     

   • parameters

     • required

     -n --name -> protein name.
     -i --input -> input path.
     

     • optional

     -c --chain -> chain name. Chain name of a FASTA file. For example, '2'. This can be empty if you prefer a name of the input FASTA file like '2wsc2' or a CASP name 'T1024'.
     

2. part 1 of feature generation via src/gdca.julia

   • description

     gdca.julia is used to generate Gaussian DCA file. You'd better run it cf. https://github.com/carlobaldassi/GaussDCA.jl.

3. prediction via run_deephelicon.py

   • description

     If you have the feature files shown in the input/ directory, you can skip over steps 1-2 to step 3. We tested this step in a rigorous way. Be sure of every feature file already in the input/ or your preferred input file path. Before start, an available EVcouplings tool must be configured with our program properly. Finally, it works easily like this.

   • python commands

     • general

     python run_deephelicon.py -n NAME -c CHAIN -i /YOUR/input/PATH/ -o /YOUR/OUTPUT/PATH/ -f FORMAT
     

     • example

     python run_deephelicon.py -n 2wsc -c 2 -i ./input/ -o ./output/ -f 'Normal'
     

   • parameters

     • required

     -n --name -> protein name.
     -i --input -> input path.
     -o --output --> prediction results.
     -f --format --> Format of a output file, 'Normal' or 'CASP14'.
     

     • optional

     -c --chain -> chain name. Chain name of a FASTA file. For example, '2'. This can be empty if you prefer a name of the input FASTA file like '2wsc2' or a CASP name 'T1024'.
     

     • see detail

     python run_deephelicon.py -h
     

4. description of an output file

   • format This predictor returns predictions of inter-helical residue contacts in tansmembrane proteins. If non-transmembrane segment or <1 transmembrane segment is detected, the programe will not return final results. However, you can still utilize the intermediate results at stage 1 and 2 as stated in the paper. Considering <1 helix detection by inside transmembrane topology predictor, we will consider extending our module to generate a file including entire results in the future work. The DeepHelicon provides two formats of output file illustrated by the two examples below (see one file format of a output file ./output/2wsc2.deephelicon).

   • example in 'Normal' format

     

   • example in 'CASP14' format

     see CASP14 format

How to cite

Sun, J., Frishman, D. (2020). DeepHelicon: accurate prediction of inter-helical residue contacts in transmembrane proteins by residual neural networks. J. Struct. Biol., vol. 212. 107574. doi:10.1016/j.jsb.2020.107574

Contact

If you have any problem in using it, please feel free to contact

jianfeng.sunmt{[({at})]}gmail.com jianfeng.sun{[({at})]}tum.de