RifDock Setup Protocol for Linux

This is the install guide to build all requirements for RifDock and RifDock itself.
Tested on Red Hat 8.3.1 CentOS 8
If you want to directly see the method to build RifDock, see the step #8
All copyright of Rifdock is on Longxing Cao and Brian Coventry et al., who developed and shared Rifdock.
The original resources of Rifdock are available from https://github.com/rifdock/rifdock
The research paper of Rifdock is available from https://www.nature.com/articles/s41586-022-04654-9
I made simple wiki for this page: https://github.com/srgokribb/Rifdock-Total-Setup/wiki

Least spec to run rifdock

Recommended by authors of Rifdock: 32+ core CPU, 16-80GB RAM depending on the size of the target
As for installation, ~10 cpus were enough.

Programs and Resources required to run RifDock

(Bolded programs are directly necessary to build Rifdock)

Rosetta_3.13 (Recent Rosetta later than 2020, HDF5 is needed)
DAIphaBall (Part of Rosetta)
PyRosetta4
PatchDock
PsiPred
Environment modules 4.5.2 (It is necessary to load another version of GCC.)
GCC-6.5.0 (Necessary to build and run rifdock)
Rosetta_3.9 compiled as cxx11_omp (Re2c and Ninja are needed)
RifDock (Boost 1.65.0 is needed)
silent_tools
Scaffolds
ss_grouped_vall_all.h5 (optional)

Notice

In this protocol, the command line which starts with "#" means "root" or "administer", and the command line which starts with "$" means "local user".
Caution: Do not install multiple versions of compilers or libraries under usr/local. If multiple versions of same compiler are accessible to $PATH or $LD_LIBRARY_PATH, the compiler will not work properly and could make fatal error. If you are an user of the server and new to linux, be cautious and ask a help to the system manager or expert.

Directory Structure

    home/   root/    usr/   opt/

    /root/packages/
                    hdf/     boost-1.65.0/    

    /home/users/local_user/packages/
                     ninja-1.10.2/    re2c-1.0.3/   gcc-6.5.0/    gcc/

    /home/users/local_user/rosetta/
             rosetta_src_release_3.9/     PyRosetta4/      rosetta_src_release_3.13/    rifdock/    scaffolds/    
             psipred/     PatchDock/       ppi_tools/       silent_tools/

    /usr/local/
             bin/   lib/  include/  share/
    
    /usr/share/Modules/
                    modulefiles/gcc/6.5.0    <-- This "6.5.0" file is the modules file which could be loaded by Environment modules

These are the major directories which are involved in setting up rifdock

1. Install HDF5 and Compile Rosetta_3.13 with HDF5

1) Install HDF5 (Ref: https://fossies.org/linux/hdf5/release_docs/INSTALL)

(1) Download HDF5 source file from The HDF5 Group server(https://www.hdfgroup.org/downloads/hdf5) . - Install ver 1.12.1.

(2) Login as administer and make the "packages" directory (This directory will be used to install new libraries, compilers, and etc).

# mkdir packages
# cd packages          // pwd is /root/packages/

This is just an example, but I recommend you to make your own directory to install compilers or programs.

(3) Move the downloaded file to the directory you've made (packages/) and unpack the tar file.

# tar -xvzf hdf5-1.12.1-centos8_64.tar.gz

If you unpack the tar file, "hdf" directory will be generated.

(3) Enter in to the unpacked directory and execute installer file(.sh).

# cd hdf
# ./HDF5-1.12.1-Linux.sh
There will be some explanation and notification for the lisence.
Keep pressing "Enter" to scroll down.
If you see the questions[y/n], type "y" and press "Enter".
# ls
# cd HDF_Group/HDF5/1.12.1
# ls
bin/ include/ lib/ share/

You can see the new directory named as "hdf" is generated. At the lowest location of the directory, there are "bin", "include", "lib", "and "Share".
"bin" directory: Executables
"include" directory: Header files
"lib" directory: Library files and linkers(.so)

(4) Copy the files to proper directory. /usr/local/bin/ user/local/include/ usr/local/lib/

# cd bin
# cp -r * /usr/local/bin
# cd ../include
# cp -r * /usr/local/include
# cd ../lib
# cp -r * /usr/local/lib

Caution: If there are some files which could be overwritten, backup the previous files to the other directory and continue to the next step.

(5) Refresh the shared library cache with "ldconfig".

# ldconfig

If your computer have the system which refreshes the shared library cache, this step could be skipped.
If there are some jobs which were already running and should be maintained, it is safer to execute this command after the jobs are finished.
ldconfig command only could be executed by administer.

(6) Login as local user and set the PATH and LD_LIBRARY_PATH in .bashrc

$ cd ~
$ vi .bashrc

Add the address of /usr/local/bin to PATH, and add /usr/local/include and /usr/local/lib to LD_LIBRARY_PATH.

      # .bashrc

      # Source global definitions
      if [ -f /etc/bashrc ]; then
          . /etc/bashrc
      fi

      # User specific environment
      PATH="$HOME/.local/bin:$HOME/bin:$PATH"
      export PATH
      PATH=$PATH:/usr/local/bin
      export PATH

      LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/lib:/usr/local/include
      export LD_LIBRARY_PATH

(7) Apply the changed environment.

$ source .bashrc

After you execute this command, close the terminal and open a new terminal to continue.

(8) Check whether $PATH and $LD_LIBRARY_PATH are updated.

$ echo $PATH
$ echo $LD_LIBRARY_PATH

If PATH is updated correctly, you can proceed to rosetta installation.
Because the addresses are added to ".bashrc", current $PATh and $LD_LIBRARY_PATH will be maintained after you close your terminal or open a new terminal.

2) Install Rosetta 3.13 and compile with HDF5

Check whether the version of compilers in your machine and environment supports Rosetta before the installation (https://new.rosettacommons.org/docs/latest/build_documentation/Cxx11Support).
GCC/g++: Version 4.8 or later (https://gcc.gnu.org/releases.html)
Clang/llvm on Linux: Version 3.3 or later (https://releases.llvm.org/download.html)

(1) Go to Rosetta commons and apply for academic license (https://www.rosettacommons.org/software/license-and-download).

(2) After login in with the ID and password of the license, go to 'Downloads' and enter Rosetta 3.13 - Download Rosetta 3.13.

(3) Download Rosetta 3.13 source (5.2G) file (rosetta_src_3.13_bundle.tgz) and copy the source file to the directory on the Linux computer where the Rosetta 3.13 will be installed (https://www.rosettacommons.org/downloads/academic/3.13/).

(4) Unpack the downloaded file.

 $ tar -xvzf rosetta_src_3.13_bundle.tgz

(5) Move to 'source' directory.

$ cd rosetta_src_release_bundle/main/source

(6) Check 'scons.py' is in the source directory. This file is the software needed to compile Rosetta (already included in the rosetta bundle in main/source directory).

(7) Compile rosetta with HDF5.

$ ./scons.py -j10 mode=release extras=hdf5 rosetta_scripts

'-j 10' means using 10 cores of CPU for compiling
'mode=release' means compile with optimizations to produce a faster version of rosetta
'mode=debug' or not mentioning any mode include additional checks which slow down Rosetta runs (not recommended).
'extra=hdf5' option is used to build basic rosetta. If you use this option, add 'rosetta_scripts' at the end of the line.
'extras=static' builds static binaries which could be useful to copy or run the apps on other systems.
'extras=mpi' compiles Rosetta in Massage Passing Interface format which supports MPI run (Recommended for the cluster system like HPC)
If the extra option is mpi, use "bin/rosetta_scripts.mpi.linuxgccrelease" instead of "rosetta_scripts". If you chose to build as static, then use "bin/rosetta_scripts.default.linuxgccrelease".
There are additional option to compile with specific cxx version. In that case, use the form like "./scons.py -j 10 mode=release bin cxx=clang cxx_ver=4.5".
The compiling process will take about 1h. The time could be reduced by increasing the number of CPUs ('-j 20' or more)

(8) Check the compiled result. If the compilation was successfully done, the last sentence of the terminal shows the result like below.

"Install file: "build/src/release/linux/4.18/64/x86/gcc/8/default/rosetta_scripts.default.linuxgccrelease" as "bin/rosetta_scripts.default.linuxgccrelease"
scons: done building targets.

(9) The executable rosetta which is named as "rosetta_scripts.hdf5.linuxgccrelease" will be in /main/source/bin.

2. Install DAIphaBall

(1) DAIphaBall is a part of Rosetta. Move to rosetta_src_release/main/source/external/DAIphaBall.

(2) Type "make" to build DAIphaBall.

$ make

DAlphaBall.gcc will be made.

3. Install PyRosetta

(1) Get the license for PyRosetta in Rosetta Commons(https://www.rosettacommons.org/software/license-and-download)

(2) Check the Python version installed on the Linux computer.

$ python --version

(3) Download proper version of PyRosetta4 in PyRosetta server(https://graylab.jhu.edu/download/PyRosetta4/archive/release/)

Release : Speed optimized build for production runs.
MinSizeRel : Build optimized to reduce memory which could be suitable for low memory systems.
Debug : Binaries compiled in debug mode with additional asserts enabled and with debug-info compiled in. Build for debugging
Python-x.x versions : The builds of PyRosetta for the specific Python version.
"PyRosetta4.Release.python36.linux.release-316.tar.bz2" is selected for installation.

(4) Move the downloaded file to the directory to install PyRosetta and unzip the file.

$ tar -vjxf PyRosetta4.Release.python36.linux.release-316.tar.bz2

(5) Enter to PyRosetta directory and type

$ cd setup
$ python setup.py install

This step requires administrative access. You need to sign in as administer or "root" account.

(6) Check whether the PyRosetta is successfully installed. If it was successful, you can see these sentences in the last sentence of the installation.

Installed /usr/local/lib/python3.6/site-package/pyrosetta-2022.14+release.d95c942-py3.6-linux-x86_64.egg
Processing dependencies for pyrosetta==2022.14+release.d95c942
Finished processing dependencies for pyrosetta==2022.14+release.d95c942

4. Install PatchDock

(1) Download PatchDock from PatchDock server(https://bioinfo3d.cs.tau.ac.il/PatchDock/).

(2) Unzip the file by

$ unzip patch_dock_download.zip

(3) PatchDock do not require further install process. Read 'README.md' to run PatchDock.

5. Install PsiPred

PSIPRED Version 4.0 By David Jones, January 2016

(1) Download PsiPred source code from Git-hub pirspred/pirspred (https://github.com/psipred/psipred).

The file name will be "psipred-master.zip".

(2) Move the zip file to the directory to install psipred in the Linux computer.

(3) Unpack the zip file by

$ unzip psipred-master.zip
$ mv psipred-master pripred

(4) Install psipred.

$ cd psipred
$ cd src
$ make
$ make install
$ export PATH=$PATH:/path/to/psipred

If the installation is finished, executable "psipred" will be in bin directory
You can use "runpsipred_single" to run psipred program.

(5) Open "runpsipre_single" and revise the path to the "execdir" and "datadir".

$ cd psipred
$ vi runpsipred_single

    #!/bin/tcsh

    # This is a simple script which will carry out all of the basic steps
    # required to make a PSIPRED prediction. Note that it assumes that the
    # following programs are available in the appropriate directories:
    # seq2mtx - PSIPRED V4 program
    # psipred - PSIPRED V4 program
    # psipass2 - PSIPRED V4 program

    # NOTE: Script modified to be more cluster friendly (DTJ April 2008)

    # Where the PSIPRED V4 programs have been installed
    set execdir = ./bin                                               // Change here to /path/to/psipred/bin

    # Where the PSIPRED V4 data files have been  installed
    set datadir = ./data                                              // Change here to /path/to/psipred/data

(6) Test whether the psipred is run

$ cd example
$ ../runpsipred_single example.fasta

Rifdock protocol only requires "runpsipred_single" executable file, so there is no need to install PSI-Blast, NCBI Toolkit, UNIREF90 and etc.

6. Set the environment to install and run rifdock (Environment modules, gcc-6.5.0)

1) Install Environment modules

(1) Check whether your computer have Environment Modules.

[Caution] Before you install Environment modules, it is critical to check whether there is already installed Environment modules. If you install additional module, the previous path settings of modules will be corrupted and it will cause a lot of fatal problems.
You can check whether you have moduels by below commands.

$ module --version
$ which module
$ whereis Module

(2) Download source files of the modules (https://sourceforge.net/projects/modules/files/Modules/modules-4.5.2/).

(3) Install Environment modules (Ref: https://www.admin-magazine.com/HPC/Articles/Environment-Modules).

# cd /usr/local
# mkdir Modules
# mkdir src
# cp modules-4.5.2.tar.gz /usr/local/Modules/src
# tar -xvzf modules-4.5.2.tar.gz
# cd modules-4.5.2
# ./configure
# make
# make install

2) Install GCC 6.5.0

(1) Current GCC verison could be identified by using

 $ gcc --version

The version of other compilers also could be identified by using '--version'.

(2) Check the kernel version of your system.

 $ uname -r

You need to check whether gcc-6.5.0 is compatible with the kernel version of your computer.

(3) Download GCC 6.5.0 from GNU server(https://ftp.gnu.org/gnu/gcc/gcc-6.5.0/).

(4) Move the tar.gz file to "packages" directory and unpack the file..

 $ tar -xvzf gcc-6.5.0.tar.gz

(5) Install GCC-6.5.0. on the local directory (Ref: https://gcc.gnu.org/wiki/InstallingGCC).

 $ cd gcc-6.5.0
 $ ./contrib/download_prerequisites
 $ mkdir objdir
 $ cd objdir
 $../configure --prefix=$HOME/path/to/make/GCC-6.5.0 --enable-languages=c,c++,fortran --disable-multilib
 $ make
 $ make install
 $ cd ~ && vi .bashrc

"--disable-multilib" is used not to make libraries for 32-bit, bacause my linux environment is 64-bit. Check whether the options are fit to your environment.

(6) Make "modulefile" for new gcc in /user/share/Modules/modulefiles.

 # cd /usr/share/Modules/modulefiles              //Go to the directory where Modules is installed.
 # mkdir gcc
 # cd gcc
 # vi 6.5.0                                       //This file have basically same format of "modules" file in modulefiles directory.
 
    #%Module 1.0
    #set    BASEDIR         /opt/[module-info name]
    set     BASEDIR         /path/to/new/gcc
    prepend-path    PATH            ${BASEDIR}/bin
    prepend-path    CPATH           ${BASEDIR}/include
    prepend-path    MANPATH         ${BASEDIR}/share/man
    prepend-path    LD_LIBRARY_PATH ${BASEDIR}/lib64

 # module avail                             //If you check the available modules, then you can see the modulefile of gcc is updated.
 gcc/6.5.0

(7) Login as local user and load new gcc using environment modules.

 $ module load gcc/6.5.0
 
If you want to automatically load gcc-6.5.0, open .bashrc and add "module load gcc/6.5.0" at the end of the file.

 $ vi .bashrc
 
    if [ -f /etc/bashrc ]; then
            . /etc/bashrc
    fi

    # User specific environment
    PATH="$HOME/.local/bin:$HOME/bin:$PATH"
    export PATH
    PATH=$PATH:/home/users/srgo/packages/re2c-1.0.3/build:/home/users/srgo/packages/ninja-1.10.2
    export PATH

    # User specific aliases and functions
    module load gcc/6.5.0

If you don't want to load gcc-6.5.0, just add "#" infront of "module load gcc/6.5.0" and inactivate it.
You need to load gcc-6.5.0 when you build or run rifdock.
If you want to make other users to accessible to gcc-6.5.0, you need to change the authority of your directory by using "chmod". Making the system which load gcc-6.5.0 from /opt/gcc/gcc-6.5.0 could be another option (I used this method, but try to build on your local directory first).

(8) Check whether the current GCC version is changed to gcc-6.5.0

$ which gcc
~/path/to/gcc-6.5.0/bin/gcc

7. Install Boost and Ninja to build RifDock

To build RifDock, obtain a copy of gcc with version >= 5.0
Install Boost version 1.65 or later
Tested CentOS8(Red Hat gcc 8.3.1) with Boost-1.65.0

1) Install Boost

Notification: You need to log in as "root" or "administer" to the properly install boost.

(1) Download Boost version 1.65.0 from Boost C++ Library server(https://www.boost.org/users/history/version_1_65_0.html).

Latest version of Boost (1.74.0 or 1.78.0) was not compatible with Rifdock.

(2) Move the downloaded file to the directory to install and unpack the file (Ref: https://valelab4.ucsf.edu/svn/3rdpartypublic/boost/more/getting_started/windows.html).

 # cd /root/
 # mkdir packages
 # cd packages
 # tar -xvzf boost_1_65_0.tar.gz
 # cd boost_1_65_0

(3) Select the location to install the boost by using

 # ./bootstrap.sh --prefix=/path/to/install/boost/

If you are sign in as "root" or administer and want to make them automatically added to /usr/local directory, you don't need to add --prefix option. Check whether libboost_xxx.so and libboost_xxx.so.1.65.0 files are generated in /usr/local/lib and /usr/local/include.

(4) Start installation

 # ./b2 install

The installation process will take a quite long time.
Even though there could be some minor errors like "warning: unnecessary parentheses in the declaration of 'assert_mot_arg' [-Wparentheses]", it didn't make any error during compiling Rifdock.
Caution: If multiple versions of boost are installed, it makes an error when compiling RifDock. So, if you want to use another version of boost, it is recommended to delete the previous boost (check /usr/local/include /usr/local/lib /usr/lib) and install a new one.

2) Load gcc-6.5.0 by modules.

- Important: From here to finishing the building rifdock, keep gcc-6.5.0 mounted on the terminals. See 6.2)-(8)

$ module load gcc/6.5.0
$ which gcc
~/packages/gcc/bin/gcc

If you don't compile rifdock using GCC 5.x, 6.x, or 7.x, you cannot install really working rifdock.
You can install rifdock even though you do not change GCC version, you will eventually get "ASSERTION ERROR" when you try to use rifgen, one of the major executables in Rifdock.

3) Install re2c - Needed to run Ninja

(1) Download re2c from downloading site (https://opensuse.pkgs.org/15.3/opensuse-oss-x86_64/re2c-1.0.3-1.18.aarch64.rpm.html).

(2) Install re2c by using following commands (Ref: https://re2c.org/build/build.html).

$ tar -xvzf re2c-1.0.3.tar.gz
$ cd re2c-1.0.3
$ mkdir build
$ cd build
$ ../configure && make && make install

(3) Add the path to executable re2c

$ export PATH=$PATH:/path/to/re2c-1.0.3/build

Add the path to the directory where "re2c" executable is installed to $PATH.

4) Install Ninja

- In order to build rosetta cxx11_omp, Ninja should be installed and proper PATH of Ninja need to be set.

(1) Download Ninja 1.10.2 from Git-hub(https://github.com/ninja-build/ninja/releases). (Ref:https://github.com/ninja-build/ninja/wiki, https://github.com/ninja-build/ninja)

(2) Install Ninja

 $ tar xvzf- ninja-1.10.2.tar.gz
 $ cd ninja-1.10.2
 $ ./configure.py --bootstrap

(3) Set the proper PATH of Ninja (Necessary)

 $ echo $PATH
 $ export PATH=$PATH:/home/users/local/ninja-1.10.2/
 $ echo $PATH

You can add new PATH by using "export PATH=$PATH:/path/to/directory"

8. Build Rosetta cxx11_omp and RifDock

Rifdock is made based on OpenMP(OMP) which is the compiler could make single thread program to multithread program.
Because Rifdock requires some libraries and resources in Rosetta, Rosetta also need to be build with OMP.

1) Build Rosetta 3.9 as cxx11_omp using Ninja

(1) Download Rosetta 3.9 source (2.8G) file (rosetta_src_3.9_bundle.tgz) and copy the source file to the directory where the Rosetta 3.9 will be installed.

(2) Unpack Rosetta

 $ tar -xvzf rosetta_src_3.9_bundle.tgz

(3) Move to main/source directory

 $ cd rosetta_src_release_bundle/main/source

(4) build rosetta cxx11_omp using Ninja(Ref: See 'Extra tools' in https://ninja-build.org/manual.html)

 $ cd rosetta_src_release_bundle/main/source
 $ CXX=/path/to/gcc-6.5.0/bin/g++ CC=/path/to/gcc-6.5.0/bin/gcc ./ninja_build.py cxx11_omp -t rosetta_scripts -remake

2) Build RifDock

(1) Copy rifdock repository and build Rifdock (https://github.com/rifdock/rifdock)

 $ unzip rifdock-master.zip
 $ mv rifdock-master rifdock
 $ cd rifdock
 $ mkdir build
 $ cd build
 $ CXX=/path/to/gcc-6.5.0/bin/g++ CC=/path/to/gcc-6.5.0/bin/gcc CMAKE_ROSETTA_PATH=/Path/to/a/rosetta/main CMAKE_FINAL_ROSETTA_PATH=/Path/to/a/rosetta/main/source/cmake/build_cxx11_omp cmake .. -DCMAKE_BUILD_TYPE=Release
 $ make -j10 rif_dock_test rifgen

When later version of Boost(Boost-1.74.0 or Boost-1.78.0) was used to build rifdock, instead of Boost-1.65.0, it caused BOOST_BITMASK error during "make" step.

2) The unit tests executable file is at:

$ rifdock/build/schemelib/test/test_libscheme

10. Download misc(cao_2021_protocol, scilent_tools, ppi_tools, scaffolds) for RifDock

1) Download cao_2021_protocol

(1) Download design scripts and main pdb files from http://files.ipd.uw.edu/pub/robust_de_novo_design_minibinders_2021/supplemental_files/scripts_and_main_pdbs.tar.gz

(2) Move the file to 'resources' directory and unpack the tar file.

$ cd resources
$ tar -xvzf scripts_and_main_pdbs.tar.gz

(3) Check the files in cao_2021_protocol.

$ cd scripts_and_main_pdbs/supplemental_files/

There will be cao_2021_protocol_guide.txt which is the practical guide to run rifdock.
There will be .xml, .py, .flags, .sh, .ipynb and etc. The usage of the files are described in cao_2021_protocol_guide.txt.

2) Download scilent_tools and ppi_tools

(1) scilent_tools and ppi_tools are in cao_2021_protocol/github_backup.

$ cd cao_2021_protocol/github_backup/

(2) Add the PATH to scilent_tools to .bashrc (This will be covered in step #11).

3) Download Miniprotein Scaffold Library

(1) Download all the scaffolds from http://files.ipd.uw.edu/pub/robust_de_novo_design_minibinders_2021/supplemental_files/scaffolds.tar.gz