Skip to content

Latest commit

 

History

History
85 lines (76 loc) · 3.94 KB

mcmc_configuration.md

File metadata and controls

85 lines (76 loc) · 3.94 KB
Raw

Setting Parameters of MCMC dimension reduction

RiD-kit uses a JSON-format file to configure MCMC dimension reduction. Here we explain these parameters one by one.

{
        "name": "rid-mcmc",
        "init_models": ["model_000.pb","model_001.pb","model_002.pb","model_003.pb"],
        
        "MCMC_Config": {
            "cv_dimension": 2,
            "numb_steps": 10000,
            "numb_walkers": 500,
            "temperature": 300,
            "proj_info": {
                "proj_mode": "cv",
                "proj_cv_index": [0,1]
            },
            "cv_type": "dih",
            "bins": 101
        }
    
},
{
        "name": "rid-mcmc",
        "init_models": ["model_000.pb","model_001.pb","model_002.pb","model_003.pb"],
        
        "MCMC_Config": {
            "cv_dimension": 20,
            "numb_steps": 100000,
            "numb_walkers": 2000,
            "cv_upper_bound": [4.5,5.0,6.0,4.2,4.0,4.0,3.5,4.0,4.0,3.5,3.2,3.0,3.5,3.5,4.0,4.5,4.5,3.75,3.8,4.0],
            "cv_lower_bound": [1.0,2.0,3.5,2.8,2.5,2.0,1.2,1.5,1.5,1.0,1.8,0.5,1.0,1.0,1.5,1.5,1.5,1.75,2.2,2.6],
            "proj_info": {
                "proj_mode": "cv",
                "proj_cv_index": [0,1]
            },
            "cv_type": "dis",
            "bins": 101
        }
    
}
  • "init_models" (List[str]) List of trained free energy models generated in the RiD run.
  • "cv_dimension" int An integer to represent the number of CVs used in RiD.
  • "numb_steps" int An integer to represent the number of steps for MCMC run.
  • "numb_walkers" int Number of parallel walkers used in the MCMC run.
  • "temperature" int Temperature used in the RiD run and MCMC simulation.
  • "cv_upper_bound" Optional[(List[float])] Upper bound of the initial value in MCMC for each dimension of CVs, this is usually set for distance CVs.
  • "cv_lower_bound" Optional[(List[float])] Lower bound of the initial value in MCMC for each dimension of CVs, this is usually set for distance CVs.
  • "proj_mode" str The mode for projecting 2D CV, support cv and path mode. cv mode representing the common projection on the selected CV index, path mode representing the projection on the user defined path CV.
  • "proj_cv_index" (List[int]) Projected index for the 2D CV. Note that 1D projection is done for all CVs.
  • "cv_type" str CV type for projection, support dih and dis, representing dihedral CVs and distance CVs.
  • "bins" int Bins for projecting CVs at each dimension.

Another example is given to project 2D path CV.

{
        "name": "rid-mcmc",
        "init_models": ["model_000.pb","model_001.pb","model_002.pb","model_003.pb"],
        
        "MCMC_Config": {
            "cv_dimension": 13,
            "numb_steps": 100000,
            "numb_walkers": 2000,
            "cv_upper_bound": [5.33507,4.43402,4.59772,3.40287,4.67268,5.12813,3.28143,3.74967,3.58144,5.34143,7.13014,4.58024,7.41678],
            "cv_lower_bound": [2.38722,0.99005,3.99886,2.08580,3.81799,3.93331,2.768175,3.145618,3.161886,4.661101,5.286805,4.076754,2.893047],
            "proj_info": {
                "proj_mode": "path",
                "proj_cv_index": [0,1,2,3],
                "path_list": [[2.38722,0.99005,3.99886,2.08580],[2.64067,1.20917,3.73831,1.86004],[2.90639,1.50430,3.47935,1.65185],
                [3.18130,1.83919,3.22240,1.46871],[3.46322,2.19574,2.96795,1.32108],[3.75057,2.56491,2.71672,1.22191],
                [4.04218,2.94198,2.46968,1.18344],[4.33720,3.32424,2.22824,1.21147],[4.63498,3.71010,1.99443,1.30170]],
                "path_lm": 0.25
            },
            "cv_type": "dis",
            "bins": 101
        }
    
}

In the path CV projection mode, the additonal parameters are:

  • "proj_cv_index" List[int] Set of CV index used to defined the path CV.
  • "path_list" List[float] List of CVs to defined the path.
  • "path_lm" float Lambda in the path CV definition.