Options for YAML files¶
These are all the simulation, alchemy, and file I/O options controlled by the options header in the YAML files for
YANK. We have subdivided the categories below, but all settings on this page go under the options header in the YAML file:
Besides the options listed in General Options that can be specified exclusively in the
options section of the YAML script, everything else can go either in options as a general setting, or in
experiments.options. In the latter case, an option can be expanded combinatorially with the !Combinatorial tag.
General Options:¶
resume_setup¶
options:
resume_setup: yes
Choose to resume a setup procedure. If no, YANK will raise an error when it detects that it will overwrite an
existing file in the directory specified by setup_dir.
Valid Options: [yes]/no
resume_simulation¶
options:
resume_simulation: yes
Choose to resume simulations. If no, YANK will raise an error when it detects that it will overwrite an existing
file in the directory specified by experiments_dir.
Valid Options: [yes]/no
output_dir¶
options:
output_dir: output
The main output folder of YANK simulations. A folder will be created if none exists. Path is relative to the YAML script path
Valid Options (output): <Path String>
setup_dir¶
options:
setup_dir: setup
The folder where all generate simulation setup files are stored. A folder will be created if none exists. Path is relative to the output_dir folder.
Valid Options (setup): <Path String>
experiments_dir¶
options:
experiments_dir: experiments
The folder where all generate simulation setup files are stored. A folder will be created if none exists. Path is relative to to the output_dir folder.
Valid Options (experiments): <Path String>
platform¶
options:
platform: fastest
The OpenMM platform used to run the calculations. The default value (fastest) automatically selects the fastest
available platform. Some platforms (especially CUDA and OpenCL) may not be available on all systems.
Valid options: [fastest]/CUDA/OpenCL/CPU/Reference
precision¶
options:
precision: auto
Floating point precision to use during the simulation. It can be set for OpenCL and CUDA platforms only. The default
value (auto) is equivalent to mixed when the device support this precision, and single otherwise.
Valid options: [auto]/double/mixed/single
max_n_contexts¶
options:
max_n_contexts: 3
The maximum number of GPU contexts that can be in memory during the simulation. In general, YANK does not need more than 3 contexts.
Valid options (3): <Integer>
switch_experiment_interval¶
options:
switch_experiments_interval: 0
When running multiple experiments using the !Combinatorial tag, this allows to switch between experiments every
switch_experiments_interval iterations, and gather data about multiple molecules/conditions before
completing the specified number_of_iterations. If 0, YANK will complete the combinatorial calculations
sequentially.
Valid options (0): <Integer>
processes_per_experiment¶
options:
processes_per_experiment: auto
When running YANK on multiple processes with MPI, you can run several experiments in parallel by using this option to
allocate a given number of processes to each experiment. This option is ignored if YANK is not run with MPI. If null,
the experiments are performed one after the other on all the available MPI processes. When auto is selected, YANK
tries to run as many experiment as possible in parallel on independent MPI processes. Currently, only
processes_per_experiment = 1 is supported for the SAMS sampler.
Valid options (auto): auto / null / <Integer>
System and Simulation Preparation:¶
randomize_ligand¶
options:
randomize_ligand: no
Randomize the position of the ligand before starting the simulation. Only works in Implicit Solvent. The ligand will be randomly rotated and displaced by a vector with magnitude proportional to randomize_ligand_sigma_multiplier with the constraint of being at a distance greater than randomize_ligand_close_cutoff from the receptor.
Valid options: [no]/yes
randomize_ligand_sigma_multiplier¶
options:
randomize_ligand_sigma_multiplier: 2.0
See randomize_ligand.
Valid options (2.0): <float>
randomize_ligand_close_cutoff¶
options:
randomize_ligand_close_cutoff: 1.5 * angstrom
See randomize_ligand.
Valid options (1.5 * angstrom): <Quantity Length> [1]
temperature¶
options:
temperature: 298 * kelvin
Temperature of the system.
Valid options (298 * kelvin): <Quantity Temperature> [1]
pressure¶
options:
pressure: 1.0 * atmosphere
Pressure of the system. If set to null, the simulation samples as an NVT ensemble.
Valid options (1 * atmosphere): null / <Quantity Pressure> [1]
hydrogen_mass¶
options:
hydrogen_mass: 1.0 * amu
Hydrogen mass for HMR simulations.
Valid options (1*amu): <Quantity Mass> [1]
constraints¶
options:
constraints: HBonds
Constrain bond lengths and angles. See OpenMM createSystem() documentation for more details.
Valid options: [Hbonds]/AllBonds/HAngles
anisotropic_dispersion_cutoff¶
options:
anisotropic_dispersion_cutoff: auto
Tell YANK to compute anisotropic dispersion corrections for long-range interactions. YANK accounts for these effects by creating two additional thermodynamic states at either end of the thermodynamic cycle with larger long-range cutoffs to remove errors introduced from treating long-range interactions as a homogeneous, equal density medium. We estimate the free energy relative to these expanded cutoff states. No simulation is actually carried out at these states but energies from simulations are evaluated at them.
This option only applies if you have specified a system with periodic boundary conditions.
We put this option in the general options category instead of the solvents section since these additional states are unique to YANK’s setup.
The size of the expanded cutoff distance can be set in a few ways through this option. If
auto the cutoff will be set to 0.99*min_box_size/2 if no barostat is in use or 0.8*min_box_size/2 if
one is in use (to account for box size fluctuations), with min_box_size denoting the norm of the smallest OpenMM
box vector defining the initial triclinic cell volume.
Valid options: [auto]/null/<Quantity Length> [1]
Simulation Parameters¶
switch_phase_interval¶
options:
switch_phase_interval: 0
This allows to switch the simulation between the two phases of the calculation every switch_phase_interval iterations.
If 0, YANK will exhaust the number_of_iterations iterations of the first phase before switching to the second one.
Valid options (0): <Integer>
minimize¶
options:
minimize: True
Minimize the input configuration before starting simulation. This is highly recommended if a pre-minimized structure is provided, or if explicit solvent generation is left to YANK. The FIRE minimizer [26], a fast minimizer that can run entirely on the GPU, is used first. If this fails, an L-BFGS minimizer [25] (as implemented in OpenMM) is used.
Valid Options: [yes]/no
minimize_max_iterations¶
options:
minimize_max_iterations: 1000
Set the maximum number of iterations the energy minimization process attempts to converge to given tolerance energy. 0 steps indicate unlimited.
Valid Options (0): <Integer>
minimize_tolerance¶
options:
minimize_tolerance: 1.0 * kilojoules_per_mole / nanometers
Set the tolerance of the energy minimization process. System is considered minimized when the energy does not change by the given tolerance in subsequent iterations.
Valid Options (1.0 * kilojoules_per_mole / nanometers): <Quantity (Molar Energy)/(Length)> [1]
number_of_equilibration_iterations¶
options:
number_of_equilibration_iterations: 1
Number of iterations used for equilibration before production run. Iterations written to file are post-equilibration.
Valid Options (1): <Integer>
equilibration_timestep¶
options:
equilibration_timestep: 1.0 * femtosecond
Timestep of the equilibration timestep (not production).
Valid Options (1.0 * femtosecond): <Quantity Time> [1]
default_number_of_iterations¶
options:
default_number_of_iterations: 1
Default number of iterations for the samplers that do not explicitly specify
the option number_of_iterations.
Note: If resume_simulation is set, this option can be used to extend previous
simulations past their original number of iterations.
Specifying 0 will run through the setup, create all the simulation files, store all options, and minimize the
initial configurations (if specified), but will not run any production simulations.
Set this to .inf (note the prepended dot character) to run an unlimited number of iterations. The simulation will
not stop unless some other criteria is stops it. We strongly recommend specifying either
online free energy analysis and/or
a phase switching interval to ensure there is at least some stop criteria,
and all phases yield some samples.
Valid Options (1): <Integer> or .inf
default_nsteps_per_iteration¶
options:
default_nsteps_per_iteration: 500
Number of timesteps between each iteration with the default MCMC move. We highly recommend using a number greater than 1 to improve decorrelation between iterations. Hamiltonian Replica Exchange swaps are attempted after each iteration. This option is ignored if a custom MCMC move is used for the experiment.
Valid Options (500): <Integer>
default_timestep¶
options:
default_timestep: 2.0 * femtosecond
The timestep of the Langevin Dynamics with the default MCMC move. This option is ignored when a custom MCMC move is used.
Valid Options (2.0 * femtosecond): <Quantity Time> [1]
checkpoint_interval¶
options:
checkpoint_interval: 50
Specify how frequently checkpoint information should be saved to file relative to iterations. YANK simulations can be
resumed only from checkpoints, so if something crashes, up to checkpoint_interval worth of iterations will be lost
and YANK will resume from the most recent checkpoint.
Note
The checkpoint also impacts disk IO times; Larger intervals consume less disk space, read faster but write slower, as a function of number of replicas. The reverse is also true. For SAMS type samplers, longer checkpoints such as 200 are fine; for replica exchange, especially in serial, lower checkpoint intervals around 10 are better. We have chosen the default of 50 to try and balance the IO and the different schemes.
This option helps control write-to-disk time and file sizes. The fewer times a checkpoint is written, the less of both
you will get. If you want to write a checkpoint every iteration, set this to 1.
Checkpoint information includes things like full coordinates and box vectors, as well as more static information such as metadata, simulation options, and serialized thermodynamic states.
Valid Options (200): <Integer >= 1>
store_solute_trajectory¶
options:
store_solute_trajectory: yes
Specify if you want an additional trajectory of just the solute atoms stored every iteration, regardless of the
checkpoint_interval.
If specified, this will write the data to the analysis file in addition to the normal information stored in the
checkpoint file. As such, you should be careful when considering space and the checkpoint_interval setting. For
instance, an implicit solvent simulation with checkpoint_interval: 1 will result in a redundant copy of the
complete trajectory.
Valid Options: [yes]/no
constraint_tolerance¶
options:
constraint_tolerance: 1.0e-6
Relative tolerance on the constraints of the system.
Valid Options (1.0e-6): <Scientific Notation Float>
Alchemy Parameters¶
annihilate_electrostatics¶
options:
annihilate_electrostatics: yes
Annihilate electrostatics rather than decouple them. This means that ligand-ligand (alchemical-alchemical) nonbonded electrostatics will be turned off as well as ligand-nonligand nonbonded electrostatics.
Valid Options: [yes]/no
annihilate_sterics¶
options:
annihilate_sterics: no
Annihilate sterics (Lennad-Jones or Halgren potential) rather than decouple them. This means that ligand-ligand
(alchemical-alchemical) nonbonded sterics will be turned off as well as ligand-nonligand nonbonded sterics.
WARNING: Do not set this option if annihilate_electrostatics is “no”.
Valid Options: [no]/yes
Steric Alchemical Options¶
options:
softcore_alpha: 0.5
softcore_a: 1
softcore_b: 1
softcore_c: 6
The options that control the soft core energy function for decoupling/annihilating steric interactions. Setting
softcore_alpha = 0 with softcore_a = 1 gives linear scaling of the Lennard-Jones energy function.
Valid Options for softcore_alpha (0.5): <Float>
Valid Options for softcore_[a,b,c] (1,1,6): <Integer preferred, Float accepted>
Electrostatic Alchemical Options¶
options:
softcore_beta: 0.0
softcore_d: 1
softcore_e: 1
softcore_f: 2
The options that control the soft core energy functnon for decoupling/annihilating electrostatic interactions.
Setting softcore_beta = 0 with softcore_d = 1 gives linear scaling of Coulomb’s law.
Valid Options for softcore_beta (0.0): <Float>
Valid Options for softcore_[d,e,f] (1,1,2): <Integer preferred, Float accepted>
alchemical_pme_treatment¶
options:
alchemical_pme_treatment: direct-space
When using PME, by default YANK runs the simulation modeling exclusively the direct space of PME. The reciprocal space is taken into account by reweighting the end states (the same reweighting performed for the anisotropic long-range dispersion correction). This makes it very fast to compute the energy matrix at each iteration. However, charged ligands may have a poor overlap between the direct-space-only and the full PME space. In this case, convergence rates can be very long, and it is recommended to use exact treatment of PME electrostatics.
Valid Options: [direct-space]/exact/coulomb
disable_alchemical_dispersion_correction¶
options:
disable_alchemical_dispersion_correction: yes
By default, the contribution of the alchemical atoms to the analytical long-range dispersion correction is not included to speed up the computation of the energy matrix. This contribution is included in the end states anisotropic long-range dispersion correction.
Valid Options: [yes]/no
split_alchemical_forces¶
options:
split_alchemical_forces: yes
By default, the alchemical forces are split into their own OpenMM force groups to speed up the computation of the energy
matrix. If your input system is particularly loaded with forces, and they occupy many force group, you may incur into
errors during the creation of the alchemical system as OpenMM supports a maximum of 32 force groups. In this case, it is
recommended to merge some of your forces into a single group. If this is not possible, set this to no to proceed
without this optimization.
Valid Options: [yes]/no
| [1] | (1, 2, 3, 4, 5, 6, 7, 8) Quantity strings are of the format: <float> * <unit> where <unit> is any valid unit specified in the “Valid Options” for an option. e.g. “<Quantity Length>” indicates any measure of length may be used for <unit> such as nanometer or angstrom.
Compound units are also parsed such as kilogram / meter**3 for density.
Only full unit names as they appear in the simtk.unit package (part of OpenMM) are allowed; so “nm” and “A” will be rejected. |