rk.refine
Run refinement with ROCKET
Running refinement with ROCKET is simple:
rk.refine config.yaml
The config.yaml
file is a human readable/editable config file with all parameters included. And rk.preprocess
will automatically generate two config files for you separately for phase1 run and phase2 run.
A typical config file is:
note: phase1_<your_note_here>
data:
datamode: xray # xray or cryoem
free_flag: R-free-flags # free label flag in your mtz file
testset_value: 1
min_resolution: 3.0 # resolution cutoff
max_resolution: null
voxel_spacing: 4.5
msa_subratio: null
w_plddt: 0.0
downsample_ratio: null # int, downsample your large cryoem mtz files
paths:
path: str # path to you working directory, should have ROCKET_inputs folder under that path
file_id: str. # system id
template_pdb: null # relative path to your template pdb if you have
input_msa: null
sub_msa_path: null
sub_delmat_path: null
msa_feat_init_path: null
starting_bias: null # Path to saved bias matrix
starting_weights: null # Path to saved weights matrix
uuid_hex: d079fd4fd9 # Automatically generately hex id for run identification
execution:
cuda_device: 0
num_of_runs: 3
verbose: false
algorithm:
bias_version: 3
iterations: 100
init_recycling: 1
domain_segs: null
optimization:
additive_learning_rate: 0.05 # Learning Rate for bias
multiplicative_learning_rate: 1.0 # Learning Rate for weights
weight_decay: 0.0001
batch_sub_ratio: 0.7
number_of_batches: 1
rbr_opt_algorithm: lbfgs
rbr_lbfgs_learning_rate: 150.0
smooth_stage_epochs: 50
phase2_final_lr: 0.001
l2_weight: 1.0e-07
features:
solvent: true # Include solvent mask contribution
sfc_scale: true # Update scales per iteration
refine_sigmaA: true # Update sigmaA per iteration
additional_chain: false
bias_from_fullmsa: false
chimera_profile: false
Check out the rk.config
functionality to also generate default config files and modify them.
Last updated