def __init__(
self,
# constants
c_localization_enty_class=LOC_ID,
c_localization_norm_class=GO_NORM_ID,
c_protein_enty_class=PRO_ID,
c_protein_norm_class=UNIPROT_NORM_ID,
c_SwissProt_organisms_used=[9606
], # TODO perhaps use other organisms ?
#
# The following two were DOCUMENT-BASED features and not domain-specific
#
f_corpus_unnormalized_total_background_loc_rels_ratios=None,
f_corpus_normalized_total_background_loc_rels_ratios=None,
#
f_SwissProt_normalized_total_absolute_loc_rels_ratios=None,
f_SwissProt_normalized_total_background_loc_rels_ratios=None,
#
f_SwissProt_normalized_exists_relation=None,
):
self.c_localization_enty_class = c_localization_enty_class
self.c_localization_norm_class = c_localization_norm_class
self.c_protein_enty_class = c_protein_enty_class
self.c_protein_norm_class = c_protein_norm_class
#
path = repo_path(
"resources", "features",
"corpus_unnormalized_total_background_loc_rels_ratios.pickle")
with open(path, "rb") as f:
self.c_corpus_unnormalized_total_background_loc_rels_ratios = pickle.load(
f)
#
# Comment so far out as it doesn't provide benefits in performance
#
# path = repo_path("resources", "features", "corpus_normalized_total_background_loc_rels_ratios.pickle")
# with open(path, "rb") as f:
# self.c_corpus_normalized_total_background_loc_rels_ratios = pickle.load(f)
#
# path = repo_path("resources", "features", "SwissProt_normalized_total_absolute_loc_rels_ratios.pickle")
# with open(path, "rb") as f:
# self.c_SwissProt_normalized_total_absolute_loc_rels_ratios = pickle.load(f)
#
# path = repo_path("resources", "features", "SwissProt_normalized_total_background_loc_rels_ratios.pickle")
# with open(path, "rb") as f:
# self.c_SwissProt_normalized_total_background_loc_rels_ratios = pickle.load(f)
self.c_SwissProt_organisms_used = c_SwissProt_organisms_used
self.c_SwissProt_all_relations = SWISSPROT_ALL_RELATIONS
#
self.f_corpus_unnormalized_total_background_loc_rels_ratios = f_corpus_unnormalized_total_background_loc_rels_ratios
self.f_corpus_normalized_total_background_loc_rels_ratios = f_corpus_normalized_total_background_loc_rels_ratios
#
self.f_SwissProt_normalized_total_absolute_loc_rels_ratios = f_SwissProt_normalized_total_absolute_loc_rels_ratios
self.f_SwissProt_normalized_total_background_loc_rels_ratios = f_SwissProt_normalized_total_background_loc_rels_ratios
#
#
self.f_SwissProt_normalized_exists_relation = f_SwissProt_normalized_exists_relation
import pickle
from nalaf.utils.download import DownloadArticle
from nalaf.utils.readers import StringReader, PMIDReader
from loctext.learning.annotators import StringTagger
from loctext.util import PRO_ID, LOC_ID, ORG_ID, REL_PRO_LOC_ID, UNIPROT_NORM_ID, GO_NORM_ID, TAXONOMY_NORM_ID, repo_path
from loctext.learning.annotators import LocTextDXModelRelationExtractor
RE_MODEL_PATH = repo_path("resources", "models",
"D0_9606,3702,4932_1497520729.163767.bin")
RE_FEATURES_PATH = repo_path(
"resources", "features", "selected",
"0_True_LinearSVC_C=2.0-1487943476.673364-NAMES.py")
RE_MODEL_BIN = None
with open(RE_MODEL_PATH, "rb") as f:
RE_MODEL_BIN = pickle.load(f)
def parse_arguments(argv=[]):
import argparse
parser = argparse.ArgumentParser(
description=
'Run LocText on some text to extract Protein<-->Cell Compartments relations'
)
input_group = parser.add_mutually_exclusive_group(required=True)
input_group.add_argument("--text", help="Run against given text/string")
input_group.add_argument(
"--pmid",
help=
from loctext.util import PRO_ID, LOC_ID, ORG_ID, REL_PRO_LOC_ID, UNIPROT_NORM_ID, GO_NORM_ID, TAXONOMY_NORM_ID, repo_path
from loctext.learning.evaluations import are_go_parent_and_child
from loctext.util import simple_parse_GO
GO_TREE = simple_parse_GO.simple_parse(
repo_path("resources", "ontologies",
"go-basic.cellular_component.latest.obo"))
Lars = [
"GO:0005576", # extracellular
"GO:0005634", # nucleus
"GO:0005739", # mitochondrion
"GO:0005764", # lysosome
"GO:0005768", # endosome
"GO:0005773", # vacuole
"GO:0005777", # peroxisome
"GO:0005783", # endoplasmic reticulum
"GO:0005794", # golgi apparatus
"GO:0005829", # cytosol
"GO:0005856", # cytoskeleton
"GO:0005886", # plasma membrane
"GO:0009507", # chloroplast
]
Tanya = [
# "GO:0016021", # integral to membrane
"GO:0009507", # chloroplast
"GO:0009535", # chloroplast thylakoid membrane
# "GO:0005737", # cytoplasm
"GO:0005783", # endoplasmic reticulum
"GO:0005789", # endoplasmic reticulum membrane
import json
from pprint import pprint
import re
# Run grep -o "GO:[0-9]*" "resources/features/human_localization_all__2016-11-20.tsv" | sort | uniq -c | sort > scripts/precomputed_SwissProt_normalized_total_absolute_loc_rels_ratio.txt
in_path = "scripts/precomputed_SwissProt_normalized_total_absolute_loc_rels_ratio.txt"
mode = "normalized"
regex_canonical_go_id = re.compile("canonical[\s\W]*?(GO:[0-9]+)")
with open(in_path) as f:
counter_relations = Counter({go: int(count) for line in f for count, go in [line.strip().split(" ")]})
GO_TREE = simple_parse_GO.simple_parse(repo_path(["resources", "ontologies", "go-basic.cellular_component.latest.obo"]))
counter_mentions = Counter()
for json_path in glob.glob("resources/features/human_localization_all_PMIDs_only_StringTagger_results__2016-11-20/*.json"):
with open(json_path) as f:
# We read the file as string since we saw json-parsing errors or inconsistencies
data = f.read()
for go in regex_canonical_go_id.findall(data):
if go in GO_TREE:
counter_mentions.update([go])
for key in counter_relations:
if key not in counter_mentions:
counter_mentions.update({key: 1})
regex_go_id = re.compile('GO:[0-9]+')
with open(in_path) as f:
next(f) # skip header
all_relations = {}
for line in f:
upid, organism_id, localization_gos = line.split("\t")
organism_id = int(organism_id)
organism_relations = all_relations.get(organism_id, {})
go_terms = regex_go_id.findall(localization_gos)
organism_relations[upid] = set(go_terms)
all_relations[organism_id] = organism_relations
print(
"Total uniprot entries:",
sum(
len(organism_relations)
for organism_relations in all_relations.values()))
print("4932 unit prot entries:", (len(all_relations[4932])))
print("9606 unit prot entries:", (len(all_relations[9606])))
print("Organisms:", set(all_relations.keys()))
out_path = repo_path("resources", "features",
"SwissProt_all_relations.pickle")
with open(out_path, "wb") as f:
pickle.dump(all_relations, f)
import pickle
from itertools import product
from loctext.util import repo_path, UNIPROT_NORM_ID, GO_NORM_ID, TAXONOMY_NORM_ID
from loctext.util import simple_parse_GO
from loctext.util.ncbi_global_align import global_align
GO_TREE = simple_parse_GO.simple_parse(
repo_path("resources", "ontologies",
"go-basic.cellular_component.latest.obo"))
"""
Dictionary with go term child --> to [list of go term parents] relationships
"""
def get_localization_name(go_id, default=""):
return GO_TREE.get(go_id, (default, "", ""))[0]
def _verify_in_ontology(term):
if term not in GO_TREE:
raise KeyError(
"The term '{}' is not recognized in the considered GO ontology hierarchy"
.format(term))
def are_go_parent_and_child(parent, child):
"""
True if terms are equal or parent is indeed a parent in the localization GO of the child. False otherwise.
"""
return _go_ids_accept_single(parent, child) is True