Loops benchmark

  • Last update: 28 November 2018
  • File size: 396.78 MB
  • Version: 1
  • Downloaded: 15
  • Author: López-Blanco J.R. and Chacón P.
5.0/5 rating 1 vote
######################################################
#  --- KORP v1.22 release - November 26th, 2018 ---  #
#  --- http://chaconlab.org/modeling/korp       ---  #
######################################################


> Loop modeling benchmarks
--------------------------

The directories named rcdXX/ contain our RCD loop modeling benchmark
(based on the SOAP-loop training set cases), where "XX" indicates 6,
8, 10, and 12 residues long loops.

Benchmark     Targets  Structures*
directory       [#]        [#]
----------------------------------
rcd6            100      100100
rcd8            100      100100
rcd10           100      100100
rcd12           100      100100
----------------------------------
* Native structures included.

Using an in-house script, 100 targets per loop length were randomly selected
from all the proteins of our PISCES training set excluding any homologous
chain (>50% sequence identity) to the 3846 proteins of SOAP-loop training set.
Then, 1000 loops were then generated by RCD for each target.

Besides the corresponding korpe (e.g. rcd6_ids.txt) and RCD (e.g. rcd6.txt)
input files, the native structures (e.g. 5FG6.pdb) and the decoys for korpe
(e.g. 5FG6_closed.pdb, 5 atoms per residue) or for SOAP-loop and Rosetta
(e.g 5K8S_decoys.txz, complete structures with all-heavy-atoms) are also
provided. In the latter tar-XZ compressed files, the side-chains were added
and repacked in PyRosetta.

Paper's results statistics can be found in the "score" file
(e.g. korp3A90c16f10_korp6no50_6Dr10n36c8b4yz8bf18_score.txt).


> Generation of more loop decoys with RCD
-----------------------------------------

The following RCD (v1.35) command generates more loop modeling decoys easily.
Note that the generated loops are randomly generated each RCD run.

#> cd rcdXX
#> rcd rcdXX.txt -r --bench -n 1000 -x ../dunbrack.bin -d 0.5 --linear -t 0.99 -o <output_directory>

Replace any "XX" in this command by 6, 8, 10, or 12 as required. Feel free
to generate more than 1000 loops changing the -n parameter.

The RCD executable and detailed info about its usage can be found here:
http://chaconlab.org/modeling/rcd


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

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

> KORP energy and benchmarks
----------------------------
Lopez-Blanco JR and Chacon P (2018). KORP: Knowledge-based 6D
potential for protein structure prediction. (to be published)

> 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.)
This email address is being protected from spambots. You need JavaScript enabled to view it.

Pablo Chacon (PhD.)
This email address is being protected from spambots. You need JavaScript enabled to view it.