-
Notifications
You must be signed in to change notification settings - Fork 0
/
west.cfg
153 lines (149 loc) · 5.82 KB
/
west.cfg
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
# The master WEST configuration file for a simulation.
# vi: set filetype=yaml :
---
west:
ddwe:
# Output directory for the run.
output_path: ddwe-logs
# Run machine learning; set to False for ablation.
do_machine_learning: True
# File containing the point for targeting. Set to None for no target.
target_point_path: common_files/reference.pdb
# Interval for plotting the latent space.
plot_interval: 10
# Drive resampling with the target in latent space (lst) or the pcoord (pcoord).
sort_by: pcoord
# Pcoord approaches zero as the target is approached.
pcoord_approaches_zero: True
# Maximum number of target seeking resamples. Effectively how many to split for lof. Merge twice as many.
max_resamples: 4
# Lower limit on walker weight. If all of the walkers are below the limit, split/merge is skipped
# that iteration
split_weight_limit: 1e-40
# Upper limit on walker weight. If all of the walkers exceed limit, split/merge is skipped
# that iteration
merge_weight_limit: 0.1
# Whether to sort by RMSD or not
not_sort_by_rmsd: False
machine_learning:
# static, train
ml_mode: static
# Contact map distance cutoff
contact_cutoff: 8.0
# How often to update the model. Set to 0 for static.
update_interval: 10
# Number of lagging iterations to use for training data.
lag_iterations: 50
# Path to the base training data (Optional).
base_training_data_path: common_files/train.npy
# Checkpoint file for static mode (Optional).
static_chk_path: static_model/model/checkpoints/checkpoint-epoch-100.pt
objective:
# Objective method to use (lof, clustering).
objective_method: lof
# Cluster resampling algorithm to use (simplified, complex).
cluster_resample_method: complex
# Function to measure distance between latent space points (cosine or euclidean).
distance_metric: euclidean_cosine
lof:
# Maximum total number of past latent space points to save for the lof scheme.
max_past_points: 1000
# Number of neighbors for LOF.
lof_n_neighbors: 20
# How many walkers to consider for splitting and merging in lof.
lof_consider_for_resample: 12
cluster:
# Clusting method (kmeans, knani, dbscan, gmm).
cluster_method: gmm
# Maximum number of contact maps to save from each cluster.
max_save_per_cluster: 40
# Number of "clusters" to use for the ablation version.
ablation_clusters: 8
# Number of KMeans clusters.
kmeans_clusters: 8
# Epsilon setting for dbscan.
dbscan_epsilon: .1
# Minimum number of points for dbscan.
dbscan_min_samples: 25
# Max components for GMM.
gmm_max_components: 8
# Threshold for outliers in GMM.
gmm_threshold: 0.8
mdance:
# Scan across multiple numbers, inclusive, e.g. 4,6
# Set to int for a single number.
knani_clusters: 3, 10
# Whether to use the Second Derivative of Davies–Bouldin index to determine minimum
db_second: true
# How often to load in structures, takes every sieve-th frame from the trajectory for analysis.
sieve: 1
# The number of frames to extract from each cluster.
n_structures: 11
# What metric to use to compare between frames, Mean-Square-Distance vs. whatever
metric: MSD
# Different cluster initialization methods. comp_sim is the k-NANI initialization
init_type: comp_sim
# Subset of data for Diversity selection. 20% * number of input structures needs to be >= the number of clusters you request (defaults to 10)
percentage: 20
drivers:
module_path: $WEST_SIM_ROOT
we_driver: deepdrive_westpa.ddmd_driver.CustomDriver
dmatrix_map: common_files/dmatrix_map.pkl
synd_model: common_files/ntl9_folding.synd
system:
driver: westpa.core.systems.WESTSystem
system_options:
# Dimensionality of your progress coordinate
pcoord_ndim: 1
# Number of data points per iteration
pcoord_len: 2
# Data type for your progress coordinate
pcoord_dtype: !!python/name:numpy.float32
bins:
type: RectilinearBinMapper
# The edges of the bins
boundaries:
- [0, 1, 'inf']
# Number walkers per bin
bin_target_counts: 36
propagation:
max_total_iterations: 2000
max_run_wallclock: 48:00:00
propagator: synd.westpa.propagator.SynMDPropagator
parameters:
synd_model: common_files/ntl9_folding.synd
topology: common_files/reference.pdb
gen_istates: false
data:
west_data_file: west.h5
datasets:
- name: pcoord
scaleoffset: 4
- name: dmatrix
scaleoffset: 4
plugins:
- plugin: common_files.augmentation_driver.SynDAugmentationDriver
module_path: $WEST_SIM_ROOT
coord_map: common_files/coord_map.pkl
dmatrix_map: common_files/dmatrix_map.pkl
# Settings for w_ipa, an interactive analysis program that can also automate analysis.
analysis:
directory: ANALYSIS # specify the directory all analysis files should exist in.
kinetics: # general options for both kinetics routines.
step_iter: 1
evolution: cumulative
extra: [ 'disable-correl' ]
analysis_schemes: # Analysis schemes. Required: name (TEST), states, and bins
TEST:
enabled: True
bins:
- type: RectilinearBinMapper
boundaries:
- [0, 1, 9.6, 'inf']
states:
- label: folded
coords:
- [0.5]
- label: unfolded
coords:
- [10]