def pdb_to_json(text, name, parser=None):
'''
Create a graph-layout displaying a pdb file which
presumably contains some RNA
The text is the contents of the pdb file.
:param text: The text of the pdb file.
:param name: The name of the pdb file.
:param parser: The PDB parser to use (Bio.PDB.PDBParser or Bio.PDB.MMCIFParser)
'''
with fus.make_temp_directory() as output_dir:
fname = op.join(output_dir, '{}.pdb'.format(name))
with open(fname, 'w') as f:
# dump the pdb text to a temporary file
f.write(text)
f.flush()
struct = parser.get_structure('temp', fname)
chains = struct.get_chains()
molecules = []
proteins = set()
rnas = set()
cgs = dict()
for chain in chains:
# create a graph json for each structure in the pdb file
if ftup.is_protein(chain):
print >> sys.stderr, "protein", chain
proteins.add(chain.id)
# process protein
molecules += [{
"type": "protein",
"header": "{}_{}".format(name, chain.id),
"seq": "",
"ss": "",
"size": len(chain.get_list()),
"uids": [uuid.uuid4().hex]
}]
pass
elif ftup.is_rna(chain):
print >> sys.stderr, "rna", chain
rnas.add(chain.id)
# process RNA molecules (hopefully)
cg = ftmc.from_pdb(fname,
chain_id=chain.id,
remove_pseudoknots=True,
parser=parser)
positions = fasta_to_positions(cg.to_fasta_string())
cg = ftmc.from_pdb(fname,
chain_id=chain.id,
remove_pseudoknots=False,
parser=parser)
cgs[chain.id] = cg
molecules += [{
"type":
"rna",
"header":
"{}_{}".format(name, chain.id),
"seq":
cg.seq,
"ss":
cg.to_dotbracket_string(),
"size":
cg.seq_length,
"uids": [uuid.uuid4().hex for i in range(cg.seq_length)],
"positions":
positions
}]
else:
# hetatm type chains which are present in MMCIF files
pass
# create a lookup table linking the id and residue number to the uid of
# that nucleotide and residue number
node_ids = dict()
for m in molecules:
for i, uid in enumerate(m['uids']):
node_ids["{}_{}".format(m['header'], i + 1)] = uid
links = []
for (a1, a2) in ftup.interchain_contacts(struct):
if (a1.parent.id[0] != ' ' or a2.parent.id[0] != ' '):
#hetatm's will be ignored for now
continue
chain1 = a1.parent.parent.id
chain2 = a2.parent.parent.id
# the source and target values below need to be reduced by the length of the
# nodes array because when the jsons are added to the graph, the link
# source and target are incremented so as to correspond to the new indeces
# of the nodes
# so a link to a node at position 10, if there are 50 nodes, will have to have
# a source value of -40
if (chain1 in proteins and chain2 in rnas):
# get the index of this nucleotide in the secondary structure
sid = cgs[chain2].seq_ids.index(a2.parent.id)
links += [{
"source":
node_ids["{}_{}_{}".format(name, chain2, sid + 1)],
"target":
node_ids["{}_{}_{}".format(name, chain1, 1)],
"link_type":
"protein_chain",
"value":
3
}]
elif (chain2 in proteins and chain1 in rnas):
# get the index of this nucleotide in the secondary structure
sid = cgs[chain1].seq_ids.index(a1.parent.id)
links += [{
"source":
node_ids["{}_{}_{}".format(name, chain1, sid + 1)],
"target":
node_ids["{}_{}_{}".format(name, chain2, 1)],
"link_type":
"protein_chain",
"value":
3
}]
elif (chain2 in rnas and chain1 in rnas):
# get the index of this nucleotide in the secondary structure
sid1 = cgs[chain1].seq_ids.index(a1.parent.id)
sid2 = cgs[chain2].seq_ids.index(a2.parent.id)
links += [{
"source":
node_ids["{}_{}_{}".format(name, chain1, sid1 + 1)],
"target":
node_ids["{}_{}_{}".format(name, chain2, sid2 + 1)],
"link_type":
"chain_chain",
"value":
3
}]
return {"molecules": molecules, "extra_links": links}
def plot_pdb(filename, ax=None):
"""
Plot a pdb file.
:param structure: A Bio.PDB.Structure
:return: An Axes object (ax)
"""
structure = Bio.PDB.PDBParser().get_structure('blah', filename)
model = list(structure)[0]
ax = None
chain_coords = {}
cgs = {}
import collections as col
# store a list of RNA nucleotides that each protein interacts with
protein_interactions = col.defaultdict(set)
protein_circles = []
for chain in model:
# iterate over RNAs
if ftup.is_rna(chain):
# convert to cg and store so that we can convert pdb nucleotide ids to
# secondary structure indexes later
cg = ftmc.from_pdb(filename, chain_id=chain.id)
cgs[chain.id] = cg
# plot the structure and store the coordinates
(ax, coords) = plot_rna(cg, offset=True, ax=ax)
chain_coords[chain.id] = coords
for (a1, a2) in ftup.interchain_contacts(structure):
# iterate over all the interactions in order to find out which
# nucleotides this protein interacts with
chain1 = a1.parent.parent
chain2 = a2.parent.parent
if ftup.is_protein(chain1) and ftup.is_rna(chain2):
# collect all the RNA nucleotides that a protein interacts with
sid = cgs[chain2.id].seq_ids.index(a2.parent.id)
protein_interactions[chain1.id].add((chain2.id, sid))
if ftup.is_rna(chain1) and ftup.is_rna(chain2):
sid1 = cgs[chain1.id].seq_ids.index(a1.parent.id)
sid2 = cgs[chain2.id].seq_ids.index(a2.parent.id)
coord1 = chain_coords[chain1.id][sid1]
coord2 = chain_coords[chain2.id][sid2]
ax.plot([coord1[0], coord2[0]], [coord1[1], coord2[1]],
'k-', alpha=0.5)
for chain in model:
# draw each protein and the links that it has to other nucleotides
if ftup.is_protein(chain):
# the protein will be positioned at the centroid of the nucleotides
# that it interacts with
interacting_coords = [np.array(chain_coords[chain_id][nuc_num])
for (chain_id, nuc_num) in protein_interactions[chain.id]]
centroid = np.sum(interacting_coords, axis=0) / len(interacting_coords)
# the size of the circle representing it will be proportional to its
# length (in nucleotides)
radius = 2 * math.sqrt(len(chain.get_list()))
protein_circles += [[centroid[0], centroid[1], radius]]
# draw all of the interactions as lines
for coord in interacting_coords:
ax.plot([coord[0], centroid[0]], [coord[1], centroid[1]], 'k-', alpha=0.5)
protein_circles = np.array(protein_circles)
if len(protein_circles) > 0:
circles(protein_circles[:,0], protein_circles[:,1], protein_circles[:,2], 'grey', alpha=0.5)
#plt.axis('off')
pass
def pdb_to_json(text, name):
'''
Create a graph-layout displaying a pdb file which
presumably contains some RNA
The text is the contents of the pdb file.
'''
with fus.make_temp_directory() as output_dir:
fname = op.join(output_dir, '{}.pdb'.format(name))
with open(fname, 'w') as f:
# dump the pdb text to a temporary file
f.write(text)
f.flush
struct = bpdb.PDBParser().get_structure('temp', fname)
chains = struct.get_chains()
jsons = []
proteins = set()
rnas = set()
cgs = dict()
for chain in chains:
# create a graph json for each structure in the pdb file
if ftup.is_protein(chain):
proteins.add(chain.id)
# process protein
jsons += [{
"nodes": [{
"group": 2,
"struct_name": "{}_{}".format(name, chain.id),
"id": 1,
"size": len(chain.get_list()),
"name": chain.id,
"node_type": "protein"
}],
"links": []
}]
pass
else:
rnas.add(chain.id)
# process RNA molecules (hopefully)
cg = ftmc.from_pdb(fname, chain_id=chain.id)
cgs[chain.id] = cg
jsons += [bg_to_json(cg)]
# create a lookup table to find out the index of each node in the
# what will eventually become the large list of nodes
counter = 0
node_ids = dict()
for j in jsons:
for n in j['nodes']:
node_ids["{}_{}".format(n['struct_name'], n['id'])] = counter
counter += 1
links = []
for (a1, a2) in ftup.interchain_contacts(struct):
if (a1.parent.id[0] != ' ' or a2.parent.id[0] != ' '):
#hetatm's will be ignored for now
continue
chain1 = a1.parent.parent.id
chain2 = a2.parent.parent.id
# the source and target values below need to be reduced by the length of the
# nodes array because when the jsons are added to the graph, the link
# source and target are incremented so as to correspond to the new indeces
# of the nodes
# so a link to a node at position 10, if there are 50 nodes, will have to have
# a source value of -40
if (chain1 in proteins and chain2 in rnas):
# get the index of this nucleotide in the secondary structure
sid = cgs[chain2].seq_ids.index(a2.parent.id)
links += [{
"source":
node_ids["{}_{}_{}".format(name, chain2, sid + 1)] -
counter,
"target":
node_ids["{}_{}_{}".format(name, chain1, 1)] - counter,
"link_type":
"protein_chain",
"value":
3
}]
elif (chain2 in proteins and chain1 in rnas):
# get the index of this nucleotide in the secondary structure
sid = cgs[chain1].seq_ids.index(a1.parent.id)
links += [{
"source":
node_ids["{}_{}_{}".format(name, chain1, sid + 1)] -
counter,
"target":
node_ids["{}_{}_{}".format(name, chain2, 1)] - counter,
"link_type":
"protein_chain",
"value":
3
}]
elif (chain2 in rnas and chain1 in rnas):
# get the index of this nucleotide in the secondary structure
sid1 = cgs[chain1].seq_ids.index(a1.parent.id)
sid2 = cgs[chain2].seq_ids.index(a2.parent.id)
links += [{
"source":
node_ids["{}_{}_{}".format(name, chain1, sid1 + 1)] -
counter,
"target":
node_ids["{}_{}_{}".format(name, chain2, sid2 + 1)] -
counter,
"link_type":
"chain_chain",
"value":
3
}]
#jsons += [{'nodes': [], "links": links}]
jsons += [{"nodes": [], "links": links}]
return {"jsons": jsons, "extra_links": links}
def test_interchain_contacts(self):
struct = bpdb.PDBParser().get_structure(
"temp", 'test/forgi/threedee/data/1MFQ.pdb')
ftup.interchain_contacts(struct)
def test_interchain_contacts(self):
struct = bpdb.PDBParser().get_structure("temp", 'test/forgi/threedee/data/1MFQ.pdb')
ftup.interchain_contacts(struct)
def test_interchain_contacts(self):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
struct = bpdb.PDBParser().get_structure(
"temp", 'test/forgi/threedee/data/1MFQ.pdb')
ftup.interchain_contacts(struct)
def plot_pdb(filename, ax=None):
"""
Plot the secondary structure of an RNA in a PDB file using the
Graph Layout from the ViennaRNA package and indicate long-range
interations and RNA-protein interactions.
Interchain RNA-RNA interactions are shown as red lines.
Proteins are shown as transparent gray circles with lines
indicating the interacting residues. The circle radius corresponds
to the number of interacting nucleotides.
:param structure: A Bio.PDB.Structure
:param ax: Optional. An matplotlib axis object
:return: An Axes object (ax)
"""
structure = Bio.PDB.PDBParser().get_structure('blah', filename)
model = list(structure)[0]
chain_coords = {}
cgs = {}
import collections as col
# store a list of RNA nucleotides that each protein interacts with
protein_interactions = col.defaultdict(set)
protein_circles = []
for chain in model:
# iterate over RNAs
if ftup.contains_rna(chain):
# convert to cg and store so that we can convert pdb nucleotide ids to
# secondary structure indexes later
cg, = ftmc.CoarseGrainRNA.from_pdb(filename, load_chains=chain.id)
cgs[chain.id] = cg
# plot the structure and store the coordinates
(ax, coords) = plot_rna(cg, offset=True, ax=ax)
chain_coords[chain.id] = coords
for (a1, a2) in ftup.interchain_contacts(structure):
# iterate over all the interactions in order to find out which
# nucleotides this protein interacts with
chain1 = a1.parent.parent
chain2 = a2.parent.parent
if ftup.is_protein(chain1) and ftup.contains_rna(chain2):
# collect all the RNA nucleotides that a protein interacts with
sid = cgs[chain2.id].seq.to_integer(fgr.RESID(chain2.id, a2.parent.id))-1
protein_interactions[chain1.id].add((chain2.id, sid))
if ftup.contains_rna(chain1) and ftup.contains_rna(chain2):
try:
sid1 = cgs[chain1.id].seq.to_integer(fgr.RESID(chain1.id, a1.parent.id))-1
sid2 = cgs[chain2.id].seq.to_integer(fgr.RESID(chain2.id, a2.parent.id))-1
except ValueError:
continue
coord1 = chain_coords[chain1.id][sid1]
coord2 = chain_coords[chain2.id][sid2]
ax.plot([coord1[0], coord2[0]], [coord1[1], coord2[1]],
'k-', alpha=0.5)
for chain in model:
# draw each protein and the links that it has to other nucleotides
if ftup.is_protein(chain):
# the protein will be positioned at the centroid of the nucleotides
# that it interacts with
interacting_coords = [np.array(chain_coords[chain_id][nuc_num])
for (chain_id, nuc_num) in protein_interactions[chain.id]]
centroid = np.sum(interacting_coords, axis=0) / \
len(interacting_coords)
# the size of the circle representing it will be proportional to its
# length (in nucleotides)
radius = 2 * math.sqrt(len(chain.get_list()))
protein_circles += [[centroid[0], centroid[1], radius]]
# draw all of the interactions as lines
for coord in interacting_coords:
ax.plot([coord[0], centroid[0]], [
coord[1], centroid[1]], 'k-', alpha=0.5)
protein_circles = np.array(protein_circles)
if len(protein_circles) > 0:
circles(protein_circles[:, 0], protein_circles[:, 1],
protein_circles[:, 2], 'grey', alpha=0.5)
# plt.axis('off')
return ax