CREST




CREST short introduction


  1. Make crest.slurm batch script for parallel calculations:

     #!/bin/bash
     #SBATCH --job-name=CREST-test
     #SBATCH --mem-per-cpu=2GB
     #SBATCH --nodes=1
     #SBATCH --ntasks=1
     #SBATCH --cpus-per-task=24
     #SBATCH -t 1-00:00:00
     #SBATCH --partition=common
    
     module load green/all
     module load xtb-crest
    
     #Run calculations 
     crest struc.xyz -gfn2 -T 24 >  struc.out
    
  2. Copy job-input file struc.xyz

  3. Submit the job on base:

        sbatch crest.slurm
    

NB! It is recommended to optimize the geometries obtained from the CREST by more accurate methods. In the end of this page are given three home made bash scripts that can be helpful during this process.




CREST long version


CREST (Conformer–Rotamer Ensemble Sampling Tool) was designed as conformer sampling program by Grimme’s group. Conformational search can be done by various levels of theory including molecular mechanics and semiempirical methods (GFNn-xTB) in gas or solvent (using several continuum models). By default CREST uses root-mean-square-deviation (RMSD) based meta-dynamics, short regular MD simulations and Genetic Z-matrix crossing (GC) algorithms. CREST can be also used for searching of protonation states, tautomerism studies and non-covalent complexes modelling. More can be found in the original article.

Environment

Environment is set up by the commands:

    module load green/all
    module load xtb-crest

Running CREST jobs

CREST input file should be in .xyz format and is executed by the command crest. This command is usually placed in slurm script.

    crest struc.xyz -gfn2 -g h2o -T 24 >  struc.out

In CREST calculation options are specified as command line arguments. Here, -T is number of processors used, -gfn2 used calculation method (here GFN2-xTB), -g h2o - GBSA implicit solvation for water. More about command line arguments and some examples of CREST commands.

Time

Time limits depend on size of a molecule and methods used, and can only be determined empirically. For eample, for a molecule of 54 atoms, 1 GB per core was enough, for 65 atoms molecule 2 GB per core were needed.

Memory

In our test runs for a molecule of 54 atoms, 1 GB per core was sufficient, but for 65 atoms molecule already 2 GB per core were needed.

How to cite:

The main publication for the CREST program - DOI: 10.1039/C9CP06869D.




Useful bash scripts


It is recommended to optimize the geometries obtained from the CREST by more accurate methods. Here are three home made bash scripts that can be helpful.

  • Start-orca.sh splits CREST output into single geometries and prepare ORCA inputs.
    NB! orca.slurm must be in the same folder as start-orca.sh and CREST calculations.
    NB! Charge, Multiplisity and Number of conformers must be given as command line argumens -c, -m and -n.

          sh start-orca.sh -c 0 -m 1 -n 500
    

    By default ORCA calculations will be done using the following method - RI-BP86-BJD3/def2-SVP . If it does not suit the method should be changed in the start-orca.sh.

  • Check.sh verifies if all calculations ended normally.

  • Crest-sorting.sh
    1. creates CREST folder and move the initial CREST calculations there
    2. merges individual ORCA optimized geometries into a shared file ALL.xyz
    3. creates a single CREST file, which then will be treated by CREST algorithms to delete double structures and sort remained structures by energy.