PconsC2

Improved contact predictions using the recognition of protein like contact patterns.

Dependencies:

• NetSurfP 1.1
• PSIPRED v3.5
• Jackhmmer from HMMER v3.0 or higher
• HHblits from HHsuite v2.0.16
• PSICOV v1.11
• plmDCA asymmetric
• either MATLAB v8.1 or higher
• or MATLAB Compiler Runtime (MCR) v8.1
• (optional: Rosetta v3.5 or weekly built)

MATLAB is needed to run plmDCA. However, if MATLAB is not available you can also use a compiled version of plmDCA. For the compiled version to run you need to provide a path to MCR.

How to run it:

Make sure all dependencies are working correctly and adjust the paths in localconfig.py.

To run PconsC2 use:

./run_pconsc2.py [-c <n_cores>] [-n n_decoys] [-m n_models]
                 [--p_psi <n_psicov_jobs>] [--p_plm <n_plmdca_jobs>]
                 [-f factor] [--norelax] [--nohoms] [--pconsc1]
                 [--rosetta]
                 <hhblits_database> <jackhmmer_database> <sequence file>

• Required:
  • hhblits_database and jackhmmer_database are paths to the databases used by HHblits and Jackhmmer
  • sequence_file is the path to the input protein sequence in FASTA format (only single sequences).
• Optional:
  • n_cores specifies the number of cores to use during computation (default: number of available cores).
  • n_decoys specifies the number of decoy structures generated by Rosetta (default: 2000).
  • n_models is the number of top-ranked models being extracted and eventually relaxed in the end (default: 10).
  • n_plmdca_jobs specifies the number of plmDCA instances run in parallel (default: min(2, n_cores)).
  • n_psicov_jobs specifies the number of PSICOV instances run in parallel (default: min(2, n_cores)).
  • factor determines the number of constraints used to fold the protein, which is: factor * length_of_the_input_sequence (default: 1.0).
  • rosetta flag to run rosetta and produce a structural model based on predicted contacts.
  • norelax is a flag that supresses relaxation of the final models. This can be used to quickly extract structures in the end (only used if rosetta flag specified).
  • nohoms is a flag that ensures that homologous structures are excluded from fragment picking. This is only useful in test cases if the model quality needs to be evaluated with a known structure (only used if rosetta flag specified).
  • --pconsc1 flag runs PconsC1 instead of PconsC2

Citation

M.J. Skwark, D. Raimondi, M. Michel, A. Elofsson. Improved contact predictions using the recognition of protein like contact patterns. PloS Computational Biology 10(11):e1003889, 2014.