def __call__(self, t):
structure_id = t[0]
if self.bio < 1:
raise ValueError('bio assembly number must be >= 1, was:',
self.bio)
# if the specified bio assembly does not exist, return an empty list
if len(t[1].bio_assembly) < self.bio:
return []
structure = ColumnarStructure(t[1])
# Get a pandas dataframe representation of the structure
df = structure.to_pandas()
if df is None:
return []
# Apply query filter
if self.query is None:
q = df
else:
q = df.query(self.query)
if q is None or q.shape[0] == 0:
return []
# Apply target filter
if self.target is None:
t = df
elif self.target == self.query:
# if query and target are identical, reuse the query dataframe
t = q
else:
t = df.query(self.target)
if t is None or t.shape[0] == 0:
return []
# Group by chain ids
q_chains = q.groupby('chain_id')
t_chains = t.groupby('chain_id')
rows = list()
# Find interactions between pairs of chains in bio assembly
transforms = self.get_transforms(structure)
for q_transform in transforms:
qindex = q_transform[0] # transformation id
qchain = q_transform[1] # chain id
if qchain in q_chains.groups.keys():
qt = q_chains.get_group(qchain).reset_index(drop=True)
else:
continue
# Stack coordinates into an nx3 array
cq = np.column_stack(
(qt['x'].values, qt['y'].values, qt['z'].values)).copy()
# Create transformation matrix
qmat = np.array(q_transform[2]).reshape((4, 4))
# Apply bio assembly transformations
# apply rotation
cqt = np.matmul(cq, qmat[0:3, 0:3])
# apply translation
cqt += qmat[3, 0:3].transpose()
for t_transform in transforms:
tindex = t_transform[0]
tchain = t_transform[1]
# exclude intra interactions (same transformation and same chain id)
if not self.intra and qindex == tindex and qchain == tchain:
continue
if not self.inter and qindex != tindex and qchain != tchain:
continue
if tchain in t_chains.groups.keys():
tt = t_chains.get_group(tchain).reset_index(drop=True)
else:
continue
# Stack coordinates into an nx3 array
ct = np.column_stack(
(tt['x'].values, tt['y'].values, tt['z'].values)).copy()
# Get a 4x4 transformation matrix
tmat = np.array(t_transform[2]).reshape((4, 4))
# Apply bio assembly transformations
# apply rotation
ctt = np.matmul(ct, tmat[0:3, 0:3])
# apply translation
ctt += tmat[3, 0:3].transpose()
rows += _calc_interactions(structure_id, qt, tt, cqt, ctt,
self.level, self.distance_cutoff,
self.bio, qindex, tindex)
return rows
def __call__(self, t):
structure_id = t[0]
# Get a pandas dataframe representation of the structure
structure = ColumnarStructure(t[1])
df = structure.to_pandas()
if df is None:
return []
# Apply query filter
if self.query is None:
q = df
else:
q = df.query(self.query)
if q is None or q.shape[0] == 0:
return []
# Apply target filter
if self.target is None:
t = df
elif self.target == self.query:
# if query and target are identical, reuse the query dataframe
t = q
else:
t = df.query(self.target)
if t is None or t.shape[0] == 0:
return []
# group by chain ids
q_chains = q.groupby('chain_id')
t_chains = t.groupby('chain_id')
rows = list()
# Find interactions between pairs of chains
for q_chain in q_chains.groups.keys():
qt = q_chains.get_group(q_chain).reset_index(drop=True)
for t_chain in t_chains.groups.keys():
# exclude intra interactions (same chain id)
if not self.intra and q_chain == t_chain:
continue
if not self.inter and q_chain != t_chain:
continue
tt = t_chains.get_group(t_chain).reset_index(drop=True)
# Stack coordinates into an nx3 array
cq = np.column_stack(
(qt['x'].values, qt['y'].values, qt['z'].values)).copy()
ct = np.column_stack(
(tt['x'].values, tt['y'].values, tt['z'].values)).copy()
rows += _calc_interactions(structure_id, qt, tt, cq, ct,
self.level, self.distance_cutoff,
None, -1, -1)
return rows