RCD Linux64

  • Last update: 28 February 2019
  • File size: 164.86 MB
  • Version: 1.40
  • Downloaded: 241
  • Author: López-Blanco JR, Canosa-Valls AJ, Li Y, and Chacón P
5.0/5 rating (2 votes)

############################################
# --- RCD v1.40 (from release v1.07)   --- #
# --- Chacon's lab - Feb 28th, 2019    --- #
############################################

The last stable version of RCD (v1.40) is released.


----------------------------------------------------
RCD (Random Coordinate Descent protein loop closure)
----------------------------------------------------

We provide statically linked 64-bit LINUX executables generated with Intel
and GNU compilers to minimize library dependence issues. Dynamic linking
was used for MPI parallelized versions (both GNU and Intel).

This is not the complete protocol for ab initio loop prediction detailed in
the paper Lopez-Blanco et al. NAR 2016 (see the complete reference below).
Please, contact us if you want to test this protocol in your local machine.

> Programs provided
-------------------
rcd v1.40 --> Protein loop closure program.

*All the programs show a basic help text using the --help option.

> Binaries releases
-------------------

Pre-compiled binaries are available in the corresponding bin/ directory. You
can choose the appropriate release for your system:

[Program] Compiler Libraries Linkage
-------------------------------------------------
*rcd_mpi Intel icpc OpenMPI dynamic
rcd_mpi_gnu GNU gcc OpenMPI dynamic
rcd Intel icpc (sequential) dynamic

* Intel compiled binaries are the fastest alternatives.


------------
REQUIREMENTS
------------

This software release has been compiled with 64-bit Intel "icpc"
(v. 19.0.2.187) and GNU "gcc" (7.3.0).

The following libraries/compilers may be required only to run the
pre-compiled binaries of the MPI release(s):

> For Intel release(s):
-----------------------

- Intel's C++ Compiler - Current tested icpc version 19.0.2.187
(gcc version 7.3.0 compatibility).

Please, download and install the Intel C++ Compiler and Math Kernel Library:
"Intel C++ Parallel Studio XE 2019 update-2 for Linux" (current tested
version is 19.0.2.187)

Intel C++ Compiler can be obtained for free from Intel's web site:
http://software.intel.com/en-us/articles/non-commercial-software-download/

- Intel's MPI implementation (v2019.2.187, it comes with Parallel Studio).

> For GNU releases:
-------------------

- GNU's "gcc" compiler (v7.3.0).

- OpenMPI v.2.1.1 (May 10, 2017)


----------------
RCD IMPROVEMENTS
----------------

The main improvements of v1.40 with respect to the previous (v1.11) are:

- Bonding Ramachandran potential implemented.
- The more accurate KORP-6D potential has replaced the old ICOSA.
- Linear exact loop closure implemented.
- Minor Ramachandran probabilites calculation issue fixed.
- Minor bugs fixed.

The main improvements with respect to the original RCD (v1.0) are:

- Fast backbone angle sampling based on neighbor-dependent Ramachandran
probability distributions.
- MPI-parallelization.
- Fast distance-orientation dependent energy filter.


------------
RCD EXAMPLES
------------

In the sub-directories rcd_seok8/ and rcd_seok12/ we have included the 8 and 12
residues long benchmarks (20 cases each) used in the RCD+ paper.

We strongly recommend to run RCD in parallel. To perform the loop closure in
parallel (with MPI) just copy and paste the following commands:

Go to the test directory:

#> cd rcd_seok12

And model the 100 lowest energy candidates from a total sampling of 1000 decoys for
20 12-residues long cases:

#> time mpirun.openmpi -np 6 ../bin/rcd_mpi_gnu seok12.txt -n 1000 -r -t 0.90 -d 0.5 --linear -x ../dunbrack.bin --energy_file ../korp6Dv1.bin --loco_best 100
--bench -o mybasename

Please, select the appropriate "mpirun" depending on the executable. Note that
OpenMPI's "mpirun.openmpi" (GNU's) was employed in this case. The total
computation time was 138 s using 12 threads in a modern Intel(R) Core(TM)
i7-8700K CPU @ 3.70GHz (6 real cores).

The same command run with Intel's version takes slightly less (128 s):

#> time mpirun -np 12 ../bin/rcd_mpi seok12.txt -n 1000 -r -t 0.90 -d 0.5 --linear -x ../dunbrack.bin --energy_file ../korp6Dv1.bin --loco_best 100 --bench -o
mybasename_intel

Despite we strongly recommend running RCD in parallel, you can also use the
sequential version of RCD ("rcd", compiled with Intel's icpc). To this end,
just remove the MPI part of the command ("mpirun.openmpi -np 6") and use
the "rcd" executable instead of "rcd_mpi_gnu".

The input file (seok12.txt) is a plain text input file. That uses the
following simple format, one loop per line:

<PDB_file> [Start_index] [End_index] [Chain_ID] [Loop_sequence]

For example:

1a8dA.pdb 155 166 A DLPDKFNAYLAN
1arbA.pdb 182 193 A WQPSGGVTEPGS
[and so on...]

The other options were those used in the recent RCD+ server paper:

-n <int>, --nloops <int>
Number of loops sampled.

-r, --randomize
Randomize bond lengths and valence angles (small perturbation).

-t <float>, --threshold <float>
Threshold for neighbor-dependent Ramachandran filter. Using 0.90
accounts for the 90% of probability. (default=0.98)

-x <string>, --pdf <string>
Dunbrack's Probability Density Function.

--energy_file <string>
Introducing a ICOSA energy data file (loco.score) enables
ICOSA energy calculations.

--loco_best <int>
Number of lowest ICOSA-energy loops that will be further stored
and processed (<nloops). (disabled by default)

-o <string>, --name <string>
Output directory basename. By default, current date will be used as
basename (e.g. run__11_May_2016__13h_23_06).

--linear
Enable the analytic linear closure method. (default= disabled)

-d <float>, --rmsd <float>
Minimum anchor RMSD distance in Angstroms (default=0.1)

--bench
Considers non-native workflow but just using native-coordinates for
RMSD benchmarking. (default = disabled)

Note that both dunbrack.bin and korp6Dv1.bin files are required to use the
neighbor-dependant Ramachandran sampling and the KORP energy filtration,
respectively. Please, provide their absolute or relative path in the command.

You can check the closed loops (*_closed.pdb file) with your favourite
molecular visualization program (e.g. VMD).

RCD outputs the following files by default:

<loop_PDB_name>_closed.pdb --> All closed loop models produced by RCD at
selected Coarse-Graining level (a single Multi-PDB file).

results.txt --> RCD's summary results file (plain text).

<loop_PDB_name>_dh/val/len.pdb --> Text files with dihedral (dh), valence
angles (val), and bond lengths (len) of all closed loops.

<loop_PDB_name>_rmsd.txt --> Text file with loop RMSD vs. native and ICOSA or
Bumps energies.

<loop_PDB_name>.log --> Log file with detailed run options data.

In case you need more help to use RCD, please, feel free to contact us!


----------
REFERENCES
----------

Please, cite our work if our tools become useful for your research.

> Improved RCD method
---------------------
Lopez-Blanco JR, Canosa-Valls AJ, Li Y, and Chacon P (2016). RCD+: Fast loop
modeling server. NAR (DOI: 10.1093/nar/gkw395).

> Original RCD method
---------------------
Chys P and Chacon P (2013). Random coordinate descent with spinor-matrices and
geometric filters for efficient loop closure. J. Chem. Theory Comput. 9:1821-1829.


-------
CONTACT
-------

Please, feel free to contact with us!
(Any suggestion or bug report is welcome)

Jose Ramon Lopez-Blanco (PhD)
jrlopez(at)iqfr.csic.es

Pablo Chacon (PhD.)
pablo(at)chaconlab.org


Structural Bioinformatics Group
IQFR-CSIC - Madrid (Spain)
(http://chaconlab.org)