mirror of
https://github.com/jertubiana/ScanNet.git
synced 2026-06-04 13:44:22 +08:00
Improved script for visualization of spatio-chemical filters
This commit is contained in:
Jérôme Tubiana
1226
Saliency analysis.ipynb
Normal file
1226
Saliency analysis.ipynb
Normal file
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
BIN
datasets/.DS_Store
vendored
BIN
datasets/.DS_Store
vendored
Binary file not shown.
@@ -7,7 +7,7 @@ import os
|
||||
from keras.models import Model
|
||||
import matplotlib
|
||||
|
||||
from preprocessing import PDBio,PDB_processing,pipelines
|
||||
from preprocessing import PDBio,PDB_processing,pipelines,protein_chemistry
|
||||
from network import neighborhoods
|
||||
import numpy as np
|
||||
|
||||
@@ -280,70 +280,7 @@ def calculate_filter_specificities(model_name,
|
||||
|
||||
|
||||
|
||||
def get_neighborhood(
|
||||
pdb = '1a3x',
|
||||
model = 0,
|
||||
chain = 'A',
|
||||
resnumber = 1,
|
||||
assembly=False,
|
||||
biounit=True,
|
||||
Kmax = 64
|
||||
|
||||
):
|
||||
filename,_ = PDBio.getPDB(pdb,biounit=biounit)
|
||||
if assembly == True:
|
||||
chain_ids = 'all'
|
||||
elif isinstance(assembly,list):
|
||||
chain_ids = assembly
|
||||
else:
|
||||
chain_ids = [(model,chain)]
|
||||
struct, chains = PDBio.load_chains(file=filename,chain_ids=chain_ids)
|
||||
pipeline = pipelines.ScanNetPipeline(aa_features='sequence',
|
||||
atom_features='type'
|
||||
)
|
||||
[aa_triplets, aa_attributes,aa_indices,aa_clouds,
|
||||
atom_triplets, atom_attributes, atom_indices, atom_clouds],_ = pipeline.process_example(chains)
|
||||
|
||||
aa_frames = neighborhoods.get_Frames(
|
||||
[[aa_triplets],[aa_clouds]], order='2')
|
||||
|
||||
|
||||
_, atom_types = neighborhoods.get_LocalNeighborhood([ [aa_frames[0]], [atom_clouds[atom_triplets[:,0]]] ],{'self_neighborhood':False,'Kmax': Kmax},attributes= [atom_attributes[atom_triplets[:,0]]],
|
||||
)
|
||||
atom_positions, atom_triplets = neighborhoods.get_LocalNeighborhood([ [aa_frames[0]], [atom_clouds[atom_triplets[:,0]]] ],{'self_neighborhood':False,'Kmax': Kmax},attributes= [atom_triplets],
|
||||
)
|
||||
|
||||
atom_types = atom_types.astype(np.int)
|
||||
atom_triplets = atom_triplets.astype(np.int)
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
resids = PDB_processing.get_PDB_indices(chains,return_chain=True,return_model=True)
|
||||
|
||||
index = np.nonzero( (resids[:,0].astype(np.int) == model) & (resids[:,1]==chain) & (resids[:,2].astype(np.int)==resnumber) )[0][0]
|
||||
|
||||
atom_positions = atom_positions[0][index]
|
||||
atom_types = atom_types[0][index]
|
||||
atom_triplets = atom_triplets[0][index]
|
||||
atom_index = list(atom_triplets[:,0])
|
||||
atom_bonds = np.zeros([Kmax,Kmax],dtype=np.bool)
|
||||
|
||||
for triplet in atom_triplets:
|
||||
if triplet[1] in atom_index:
|
||||
atom_bonds[atom_index.index(triplet[0]),atom_index.index(triplet[1])] = 1
|
||||
if triplet[2] in atom_index:
|
||||
atom_bonds[atom_index.index(triplet[0]),atom_index.index(triplet[2])] = 1
|
||||
atom_bonds += atom_bonds.T
|
||||
atom_bonds = list(zip(*np.nonzero(atom_bonds) ) )
|
||||
atom_bonds = [(i,j) for i,j in atom_bonds if j>i]
|
||||
return atom_positions,atom_types-1,atom_bonds
|
||||
|
||||
|
||||
|
||||
|
||||
def plot_atomic_filter(filter_specificities, index, threshold1=0.33, threshold2=0.25, sg=None,camera_position=None):
|
||||
def plot_atomic_filter(filter_specificities, index, threshold1=0.33, threshold2=0.25,y_offset=0,list_additional_objects = [], sg=None,camera_position=None):
|
||||
if camera_position is None:
|
||||
camera_position = [0.8, 0.5, 0.8]
|
||||
|
||||
@@ -395,7 +332,8 @@ def plot_atomic_filter(filter_specificities, index, threshold1=0.33, threshold2=
|
||||
show_frame=True,
|
||||
fs=1.,
|
||||
scale=2.0,
|
||||
offset=[0, 0, 0],
|
||||
offset=[0, y_offset, 0],
|
||||
list_additional_objects=list_additional_objects,
|
||||
camera_position=camera_position,
|
||||
key_light_position=[0.5, 1, 0],
|
||||
maxi=5)
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
import pythreejs
|
||||
import numpy as np
|
||||
from preprocessing import PDBio,PDB_processing,pipelines
|
||||
from preprocessing import PDBio,PDB_processing,pipelines,protein_chemistry
|
||||
from network import neighborhoods
|
||||
|
||||
def rgb_to_hex(rgb):
|
||||
@@ -37,7 +37,7 @@ def show_atoms(
|
||||
atom_bonds,
|
||||
render=True,
|
||||
radius_scale=0.2,
|
||||
stick_radius=0.15,
|
||||
stick_radius=0.75,
|
||||
stick_height=0.8,
|
||||
camera_position=[0.8, 0.5, 0.8],
|
||||
key_light_position=[0.5, 1, 0.0]
|
||||
@@ -78,8 +78,8 @@ def show_atoms(
|
||||
children = []
|
||||
|
||||
sphere_geometry = [pythreejs.SphereGeometry(radius=radius_scale * radius) for radius in VanDerWaalsRadii]
|
||||
stick_geometry = pythreejs.CylinderGeometry(radiusTop=stick_radius,
|
||||
radiusBottom=stick_radius,
|
||||
stick_geometry = pythreejs.CylinderGeometry(radiusTop=radius_scale *stick_radius,
|
||||
radiusBottom=radius_scale *stick_radius,
|
||||
height=stick_height)
|
||||
|
||||
list_spheres = [
|
||||
@@ -131,12 +131,21 @@ def get_neighborhood(
|
||||
model = 0,
|
||||
chain = 'A',
|
||||
resnumber = 1,
|
||||
atom = None,
|
||||
resindex = None,
|
||||
atomindex= None,
|
||||
assembly=False,
|
||||
biounit=True,
|
||||
Kmax = 64
|
||||
Kmax = None
|
||||
|
||||
):
|
||||
if Kmax is None:
|
||||
if (atom is not None) | (atomindex is not None):
|
||||
Kmax = 32
|
||||
else:
|
||||
Kmax = 16*9
|
||||
|
||||
|
||||
filename,_ = PDBio.getPDB(pdb,biounit=biounit)
|
||||
if assembly == True:
|
||||
chain_ids = 'all'
|
||||
@@ -146,37 +155,64 @@ def get_neighborhood(
|
||||
chain_ids = [(model,chain)]
|
||||
struct, chains = PDBio.load_chains(file=filename,chain_ids=chain_ids)
|
||||
pipeline = pipelines.ScanNetPipeline(aa_features='sequence',
|
||||
atom_features='type'
|
||||
atom_features='id'
|
||||
)
|
||||
|
||||
[aa_triplets, aa_attributes,aa_indices,aa_clouds,
|
||||
atom_triplets, atom_attributes, atom_indices, atom_clouds],_ = pipeline.process_example(chains)
|
||||
|
||||
aa_frames = neighborhoods.get_Frames(
|
||||
[[aa_triplets],[aa_clouds]], order='2')
|
||||
atom_triplets, atom_ids, atom_indices, atom_clouds],_ = pipeline.process_example(chains)
|
||||
atom_ids -=1
|
||||
atom_attributes = protein_chemistry.index_to_type[atom_ids]+1
|
||||
|
||||
|
||||
_, atom_types = neighborhoods.get_LocalNeighborhood([ [aa_frames[0]], [atom_clouds[atom_triplets[:,0]]] ],{'self_neighborhood':False,'Kmax': Kmax},attributes= [atom_attributes[atom_triplets[:,0]]],
|
||||
resids = PDB_processing.get_PDB_indices(chains,return_chain=True,return_model=True)
|
||||
|
||||
if (atom is not None) | (atomindex is not None):
|
||||
if (atomindex is not None):
|
||||
index = atomindex
|
||||
else:
|
||||
try:
|
||||
index = np.nonzero(
|
||||
(resids[atom_indices[:,0], 0].astype(np.int) == model) &
|
||||
(resids[atom_indices[:,0], 1] == chain) &
|
||||
(resids[atom_indices[:,0], 2].astype(np.int) == resnumber) &
|
||||
(atom_ids == protein_chemistry.atom_to_index[atom])
|
||||
)[0][0]
|
||||
except:
|
||||
raise ValueError('Atom #%s/%s:%s@%s not found' % (model,chain,resnumber,atom) )
|
||||
else:
|
||||
if resindex is not None:
|
||||
index = resindex
|
||||
else:
|
||||
try:
|
||||
index = np.nonzero( (resids[:,0].astype(np.int) == model) &
|
||||
(resids[:,1]==chain) &
|
||||
(resids[:,2].astype(np.int)==resnumber)
|
||||
)[0][0]
|
||||
except:
|
||||
raise ValueError('Residue #%s/%s:%s not found' % (model, chain, resnumber))
|
||||
|
||||
|
||||
if (atom is not None) | (atomindex is not None):
|
||||
frames = neighborhoods.get_Frames(
|
||||
[[atom_triplets[index:index+1]],[atom_clouds]], order='2')
|
||||
else:
|
||||
frames = neighborhoods.get_Frames(
|
||||
[[aa_triplets[index:index+1]],[aa_clouds]], order='2')
|
||||
|
||||
|
||||
_, atom_types = neighborhoods.get_LocalNeighborhood([ [frames[0]], [atom_clouds[atom_triplets[:,0]]] ],{'self_neighborhood':False,'Kmax': Kmax},attributes= [atom_attributes],
|
||||
)
|
||||
atom_positions, atom_triplets = neighborhoods.get_LocalNeighborhood([ [aa_frames[0]], [atom_clouds[atom_triplets[:,0]]] ],{'self_neighborhood':False,'Kmax': Kmax},attributes= [atom_triplets],
|
||||
atom_positions, atom_triplets = neighborhoods.get_LocalNeighborhood([ [frames[0]], [atom_clouds[atom_triplets[:,0]]] ],{'self_neighborhood':False,'Kmax': Kmax},attributes= [atom_triplets],
|
||||
)
|
||||
|
||||
atom_types = atom_types.astype(np.int)
|
||||
atom_triplets = atom_triplets.astype(np.int)
|
||||
|
||||
|
||||
atom_positions = atom_positions[0][0]
|
||||
atom_types = atom_types[0][0] -1 # Inside the code, atom_type = 0 corresponds to virtual atom or empty placeholder.
|
||||
atom_triplets = atom_triplets[0][0]
|
||||
|
||||
|
||||
|
||||
resids = PDB_processing.get_PDB_indices(chains,return_chain=True,return_model=True)
|
||||
|
||||
if resindex is not None:
|
||||
index = resindex
|
||||
else:
|
||||
index = np.nonzero((resids[:, 0].astype(np.int) == model) & (resids[:, 1] == chain) & (
|
||||
resids[:, 2].astype(np.int) == resnumber))[0][0]
|
||||
atom_positions = atom_positions[0][index]
|
||||
atom_types = atom_types[0][index] -1 # Inside the code, atom_type = 0 corresponds to virtual atom or empty placeholder.
|
||||
atom_triplets = atom_triplets[0][index]
|
||||
atom_index = list(atom_triplets[:,0])
|
||||
atom_bonds = np.zeros([Kmax,Kmax],dtype=np.bool)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user