def __init__(self, tf_sess_target=""):
with graph.as_default():
asset_path("encoder.index")
asset_path("encoder.meta")
p = asset_path("encoder.data-00000-of-00001")
p = os.path.join(os.path.dirname(p), "encoder")
self.sess = tf.Session(tf_sess_target)
variables = slim.get_variables_to_restore()
saver = tf.train.Saver(variables)
saver.restore(self.sess, p)
with open(asset_path("pca-coef.pickle"), "rb") as f:
# encoding='latin1' fixs the incompatibility of numpy between Python 2 and 3
# see more: https://stackoverflow.com/a/41366785
def load(f):
if sys.version_info.major == 2:
return pickle.load(f)
else:
return pickle.load(f, encoding='latin1')
_ = load(f)
_ = load(f)
_ = load(f)
_ = load(f)
_ = load(f)
_ = load(f)
self.coef = load(f)
self.pca = load(f)
self.sslst = {}
with open(asset_path("filter33.lst"), "r") as f:
for line in f:
self.sslst[line.strip()] = 1
def __init__(self, db_fn=None, cutoff=12):
""" Init the Searcher Instance. If db_fn is not indicated, it automatically download
contactlib-l4-g0-c2-d7.db as default database.
"""
if db_fn is None:
db_fn = asset_path("contactlib-l4-g0-c2-d7.db")
if os.getenv("ENV") == "Development":
cur_dir = os.path.dirname(os.path.realpath(__file__))
lib_fn = os.path.join(cur_dir, "libsearch.so")
else:
if platform.system() == "Darwin":
libsearch = "libsearch-2.1-macOS.so"
elif platform.system() == "Linux":
libsearch = "libsearch-2.1-linux-amd64.so"
else:
raise Exception("Unsupported platform! Please try it under Linux.")
lib_fn = asset_path(libsearch)
self.lib = CDLL(lib_fn)
self.lib.newDB.restype = c_uint64 # void * is 64 bit, but the defualt type is unint32
self.db = self.lib.newDB(db_fn.encode(), cutoff)
def convertPDB(pdbfn):
if platform.system() == "Darwin":
cmd = asset_path("dssp-2.0.4-macOS")
elif platform.system() == "Linux":
cmd = asset_path("dssp-2.0.4-linux-amd64")
else:
raise Exception("Unsupported platform! Please try it under Linux.")
pdbid = os.path.basename(pdbfn).replace(".pdb", "").upper()
model = PDBParser(PERMISSIVE=1).get_structure(pdbid, pdbfn)[0]
dsspfn = pdbfn.replace(".pdb", ".dssp")
subprocess.check_call([cmd, '-i', pdbfn, '-o', dsspfn])
dssp, keys = make_dssp_dict(dsspfn)
idx, res, ss = [], [], []
for k in keys:
try:
i, r, s, c = k[1][1], dssp[k][0], dssp[k][1], model[k[0]][
k[1]]["CA"].get_coord()
idx.append(i)
res.append(r)
ss.append(s)
except KeyError:
pass
fafn = pdbfn.replace(".pdb", ".fa")
with open(fafn, "w") as f:
f.write(">%s\n" % pdbid)
tmp = "".join(res)
while tmp:
if len(tmp) > 120:
f.write("%s\n" % tmp[:120])
tmp = tmp[120:]
else:
f.write("%s\n" % tmp)
tmp = ""
#! /usr/bin/env python
from __future__ import print_function, absolute_import
import sys
from contactlib.search import Searcher
from contactlib.encoder import Encoder
from contactlib.common import convertPDB
from contactlib.data_manger import asset_path
e = Encoder()
s = Searcher()
# build ContactLib fingerprint file
e.encode(asset_path("3aa0a.pdb"), "3aa0a.cl")
# scan ContactLib for homologous proteins, and store the similarity scores
s.search("3aa0a.cl", "output.txt")
#! /usr/bin/env python
from __future__ import print_function, absolute_import
import sys
from contactlib import Searcher, Encoder
from contactlib.common import convertPDB
from contactlib.data_manger import asset_path
e = Encoder()
s = Searcher()
# generate FASTA and DSSP files, given a PDB file
convertPDB(asset_path("3aa0a.pdb"))
# build ContactLib fingerprint file
e.encode(asset_path("3aa0a.pdb"), "3aa0a.cl")
# scan ContactLib for homologous proteins, and store the similarity scores
s.search("3aa0a.cl", "output.txt")