Page created 21 September 2015 by Vikram K. Mulligan, Baker laboratory (


The simple_cycpep_predict app is intended to be a fragment-free design validation tool for small, backbone-cyclized peptide designs analogous to AbinitioRelax for proteins designs. It sets up a chain with backbone-cyclization, uses GeneralizedKIC to sample conformations and close the geometry, then applies a few rounds of FastRelax, in two batches (a first with the mainchain hydrogen bond terms turned up to preserve hydrogen bonds, and a second with the mainchain hydrogen bond terms set to their normal values). User options permit users to filter based on the number of mainchain hydrogen bonds.

Typical usage

Let's say we want to generate energy landscape plots for a six-residue peptide with sequence VAL-DPRO-ARG-ILE-DPRO-GLU, and that we have a design for this peptide in the file inputs/native.pdb. Let's also suppose that we want every solution to have at least one hbond, and to be based on an initial closure with at least two hbonds. We would first generate a sequence file (let's call it sequence.txt, also in the inputs/ directory) containing the sequence "VAL DPRO ARG ILE DPRO GLU" (no quotation marks). We would then run simple_cycpep_predict on a single processor, dumping output to out.log, using the following command-line flags:

nohup <path to Rosetta bin directory>/simple_cycpep_predict.default.<os><compiler><mode> -sequence_file inputs/sequence.txt -in:file:native inputs/native.pdb -ex1 -ex2 -out:file:silent output.silent -genkic_closure_attempts 5000 -min_genkic_hbonds 2 -min_final_hbonds 1 -nstruct 1000 >out.log &

The output log will contain lines listing the RMSD, energy, and number of hbonds for each solution found. One can find these using the following command:

grep RMSD out.log -A1 | grep -v -P '\-\-|RMSD'

User options

  • nstruct [integer]: The number of jobs that will be attempted. This is the maximum number of structures that will be returned (if none of the attempts fails).
  • in::file::native [filename]: An optional PDB file for the native or design state. This must be a cyclic peptide with the same number of residues as the sequence provided. If this option is used, an RMSD value will be calculated for each job returning a structure.
  • out::file::o [prefix]: Prefix for PDB files to be written out. Default "S_".
  • out::file::silent [name]: Alternative to out::file::o. Writes all output to a single binary silent file. Good for large-scale sampling.
  • sequence_file [filename]: Required input. A text file containing the sequence, as full building block names separated by whitespace (e.g. "ALA DPRO DLYS B3F ARG PHE VAL").
  • genkic_closure_attempts [integer]: The maximum number of closure attempts that will be made for each job specified with the nstruct flag.
  • genkic_min_solution_count [integer]: The minimum number of solutions that the GeneralizedKIC algorithm must have found before the search is abandoned and the lowest-energy solution is chosen. By default, this is set to 1 (which picks a solution as soon as any are found). Setting this to 0 means that solutions will be sought until the number of attempts equals the number set with the genkic_closure_attempts flag, at which point one will be chosen.
  • cyclic permutations [boolean]: Should cyclic permutations of the sequence be generated during sampling, to minimize artefacts or biases introduced by the choice of the GeneralizedKIC closure points? Default true.
  • use_rama_filter [boolean]: Should GeneralizedKIC filters be rejected if alpha-amino acid pivot residues have poor Ramachandran scores? Default true.
  • rama_cutoff [real]: If the use_rama_filter option is used, this is the Ramachandran energy above which solutions are rejected. Defaults to 0.3 Rosetta energy units.
  • high_hbond_weight_multiplier: GenKIC solutions are chosen, and the initial rounds of FastRelax are performed, with a scoring function with the mainchain hydrogen bond term weights turned up by this factor. Default 10.
  • min_genkic_hbonds: GenKIC solutions are rejected that have fewer than this number of mainchain hydrogen bonds. Default 3.
  • min_final_hbonds: Final solutions are rejected that have fewer than this number of mainchain hydrogen bonds. Defaults to 0 (no solutions rejected).
  • count_sc_hbonds: If false, only mainchain hydrogen bonds are included in the counts when filtering GenKIC and final solutions. If true, all hydrogen bonds are counted. Default false.

Planned future features

  • Support for backbone torsion sampling of beta- and gamma-amino acids.
  • Bin transition based sampling.