|
1
|
|
|
""" |
|
2
|
|
|
Run searches and write files. |
|
3
|
|
|
""" |
|
4
|
|
|
|
|
5
|
|
|
from __future__ import annotations |
|
6
|
|
|
|
|
7
|
|
|
import gzip |
|
|
|
|
|
|
8
|
|
|
from pathlib import Path |
|
9
|
|
|
from typing import Sequence, Optional, Dict |
|
10
|
|
|
|
|
11
|
|
|
import pandas as pd |
|
|
|
|
|
|
12
|
|
|
from pocketutils.core.dot_dict import NestedDotDict |
|
|
|
|
|
|
13
|
|
|
from pocketutils.tools.common_tools import CommonTools |
|
|
|
|
|
|
14
|
|
|
from pocketutils.tools.path_tools import PathTools |
|
|
|
|
|
|
15
|
|
|
from typeddfs import TypedDfs, UntypedDf |
|
|
|
|
|
|
16
|
|
|
|
|
17
|
|
|
from mandos import logger |
|
18
|
|
|
from mandos.model import CompoundNotFoundError |
|
19
|
|
|
from mandos.model.chembl_support.chembl_utils import ChemblUtils |
|
20
|
|
|
from mandos.model.searches import Search |
|
21
|
|
|
from mandos.model.settings import MANDOS_SETTINGS |
|
22
|
|
|
from mandos.search.chembl import ChemblSearch |
|
23
|
|
|
from mandos.search.pubchem import PubchemSearch |
|
24
|
|
|
from mandos.entries.api_singletons import Apis |
|
25
|
|
|
|
|
26
|
|
|
InputFrame = (TypedDfs.typed("InputFrame").require("inchikey")).build() |
|
27
|
|
|
|
|
28
|
|
|
IdMatchFrame = ( |
|
29
|
|
|
TypedDfs.typed("IdMatchFrame") |
|
30
|
|
|
.require("inchikey") |
|
31
|
|
|
.require("chembl_id") |
|
32
|
|
|
.require("pubchem_id") |
|
33
|
|
|
.strict() |
|
34
|
|
|
).build() |
|
35
|
|
|
|
|
36
|
|
|
|
|
37
|
|
|
class SearcherUtils: |
|
|
|
|
|
|
38
|
|
|
@classmethod |
|
39
|
|
|
def dl( |
|
|
|
|
|
|
40
|
|
|
cls, |
|
|
|
|
|
|
41
|
|
|
inchikeys: Sequence[str], |
|
|
|
|
|
|
42
|
|
|
pubchem: bool = True, |
|
|
|
|
|
|
43
|
|
|
chembl: bool = True, |
|
|
|
|
|
|
44
|
|
|
hmdb: bool = True, |
|
|
|
|
|
|
45
|
|
|
) -> IdMatchFrame: |
|
46
|
|
|
# we actually cache the results, even though the underlying APIs cache |
|
47
|
|
|
# the reasons for this are a little obscure -- |
|
48
|
|
|
# when running a Searcher, we want to run before the FIRST search |
|
49
|
|
|
# for the typer commands to be replicas of the ``Entry.run`` methods, Searcher fetches before running a search |
|
|
|
|
|
|
50
|
|
|
# but if we have multiple searches (as in ``mandos search --config``), we only want that at the beginning |
|
|
|
|
|
|
51
|
|
|
# the alternative was having ``mandos search`` dynamically subclass each ``Entry`` -- which was really hard |
|
|
|
|
|
|
52
|
|
|
# this is much cleaner, even though it's redundant |
|
53
|
|
|
# if the cached results under /pubchem and /chembl are deleted, we unfortunately won't cache the results |
|
|
|
|
|
|
54
|
|
|
# when running this command |
|
55
|
|
|
# to fix that, we need to delete the cached /match dataframes |
|
56
|
|
|
# now that I'm writing this down, I realize this is pretty bad |
|
57
|
|
|
# TODO |
|
|
|
|
|
|
58
|
|
|
# noinspection PyPep8Naming |
|
59
|
|
|
Chembl, Pubchem = Apis.Chembl, Apis.Pubchem |
|
|
|
|
|
|
60
|
|
|
logger.notice(f"Using {Chembl}, {Pubchem}") |
|
61
|
|
|
key = hash(",".join(inchikeys)) |
|
62
|
|
|
cached_path = (MANDOS_SETTINGS.match_cache_path / str(key)).with_suffix(".feather") |
|
63
|
|
|
if cached_path.exists(): |
|
64
|
|
|
logger.info(f"Found ID matching results at {cached_path}") |
|
65
|
|
|
return IdMatchFrame.read_feather(cached_path) |
|
66
|
|
|
found_chembl: Dict[str, str] = {} |
|
67
|
|
|
found_pubchem: Dict[str, str] = {} |
|
68
|
|
|
if pubchem: |
|
69
|
|
|
for inchikey in inchikeys: |
|
70
|
|
|
try: |
|
71
|
|
|
cid = Pubchem.fetch_data(inchikey).cid |
|
72
|
|
|
found_pubchem[inchikey] = str(cid) |
|
73
|
|
|
logger.info(f"Found: PubChem {inchikey} ({cid})") |
|
74
|
|
|
except CompoundNotFoundError: |
|
75
|
|
|
logger.info(f"NOT FOUND: PubChem {inchikey}") |
|
76
|
|
|
logger.debug(f"Did not find PubChem {inchikey}", exc_info=True) |
|
77
|
|
|
if chembl: |
|
78
|
|
|
for inchikey in inchikeys: |
|
79
|
|
|
try: |
|
80
|
|
|
chid = ChemblUtils(Chembl).get_compound(inchikey).chid |
|
81
|
|
|
found_chembl[inchikey] = chid |
|
82
|
|
|
logger.info(f"Found: ChEMBL {inchikey} ({chid})") |
|
83
|
|
|
except CompoundNotFoundError: |
|
84
|
|
|
logger.info(f"NOT FOUND: ChEMBL {inchikey}") |
|
85
|
|
|
logger.debug(f"Did not find ChEMBL {inchikey}", exc_info=True) |
|
86
|
|
|
df = pd.DataFrame([pd.Series(dict(inchikey=c)) for c in inchikeys]) |
|
|
|
|
|
|
87
|
|
|
df["chembl_id"] = df["inchikey"].map(found_chembl.get) |
|
88
|
|
|
df["pubchem_id"] = df["inchikey"].map(found_pubchem.get) |
|
89
|
|
|
df = IdMatchFrame(df) |
|
|
|
|
|
|
90
|
|
|
df.to_feather(cached_path) |
|
91
|
|
|
logger.info(f"Wrote {cached_path}") |
|
92
|
|
|
|
|
93
|
|
|
@classmethod |
|
94
|
|
|
def read(cls, input_path: Path) -> InputFrame: |
|
|
|
|
|
|
95
|
|
|
df: UntypedDf = TypedDfs.untyped("Input").read_file(input_path, header=None, comment="#") |
|
|
|
|
|
|
96
|
|
|
if "inchikey" in df.column_names(): |
|
97
|
|
|
df = InputFrame.convert(df) |
|
|
|
|
|
|
98
|
|
|
elif ".lines" in input_path.name or ".txt" in input_path.name: |
|
99
|
|
|
df.columns = ["inchikey"] |
|
100
|
|
|
df = InputFrame.convert(df) |
|
|
|
|
|
|
101
|
|
|
else: |
|
102
|
|
|
raise ValueError(f"Could not parse {input_path}; no column 'inchikey'") |
|
103
|
|
|
# find duplicates |
|
104
|
|
|
# in hindsight, this wasn't worth the amount of code |
|
105
|
|
|
n0 = len(df) |
|
|
|
|
|
|
106
|
|
|
# noinspection PyTypeChecker |
|
107
|
|
|
df: UntypedDf = df.drop_duplicates() |
|
|
|
|
|
|
108
|
|
|
n1 = len(df) |
|
|
|
|
|
|
109
|
|
|
logger.info("Read {n1} input compounds") |
|
110
|
|
|
if n0 == n1: |
|
111
|
|
|
logger.info(f"There were no duplicate rows") |
|
|
|
|
|
|
112
|
|
|
else: |
|
113
|
|
|
logger.info(f"Dropped {n1-n0} duplicated rows") |
|
114
|
|
|
duplicated = df[df.duplicated("inchikey", keep=False)] |
|
115
|
|
|
duplicated_inchikeys = set(duplicated["inchikey"]) |
|
116
|
|
|
# noinspection PyTypeChecker |
|
117
|
|
|
df = df.drop_duplicates(subset=["inchikey"], keep="first") |
|
|
|
|
|
|
118
|
|
|
n2 = len(df) |
|
|
|
|
|
|
119
|
|
|
if len(duplicated) > 1: |
|
120
|
|
|
logger.error( |
|
121
|
|
|
f"{len(duplicated)} rows contain the same inchikey but have differences in other columns" |
|
|
|
|
|
|
122
|
|
|
) |
|
123
|
|
|
logger.error(f"Dropped {n2-n1} rows with duplicate inchikeys") |
|
124
|
|
|
logger.error(f"The offending inchikeys are {duplicated_inchikeys}") |
|
125
|
|
|
return df |
|
126
|
|
|
|
|
127
|
|
|
|
|
128
|
|
|
class Searcher: |
|
129
|
|
|
""" |
|
130
|
|
|
Executes one or more searches and saves the results to CSV files. |
|
131
|
|
|
Create and use once. |
|
132
|
|
|
""" |
|
133
|
|
|
|
|
134
|
|
|
def __init__(self, searches: Sequence[Search], to: Sequence[Path], input_path: Path): |
|
135
|
|
|
""" |
|
136
|
|
|
Constructor. |
|
137
|
|
|
|
|
138
|
|
|
Args: |
|
139
|
|
|
searches: |
|
140
|
|
|
input_path: Path to the input file of one of the formats: |
|
141
|
|
|
- .txt containing one InChI Key per line |
|
142
|
|
|
- .csv, .tsv, .tab, csv.gz, .tsv.gz, .tab.gz, or .feather containing a column called inchikey |
|
|
|
|
|
|
143
|
|
|
""" |
|
144
|
|
|
self.what = searches |
|
145
|
|
|
self.input_path: Optional[Path] = input_path |
|
146
|
|
|
self.input_df: InputFrame = None |
|
147
|
|
|
self.output_paths = { |
|
148
|
|
|
what.key: self._output_path_of(what, path) |
|
149
|
|
|
for what, path in CommonTools.zip_list(searches, to) |
|
150
|
|
|
} |
|
151
|
|
|
|
|
152
|
|
|
def search(self) -> Searcher: |
|
153
|
|
|
""" |
|
154
|
|
|
Performs the search, and writes data. |
|
155
|
|
|
""" |
|
156
|
|
|
if self.input_df is not None: |
|
157
|
|
|
raise ValueError(f"Already ran a search") |
|
|
|
|
|
|
158
|
|
|
self.input_df = SearcherUtils.read(self.input_path) |
|
159
|
|
|
inchikeys = self.input_df["inchikey"].unique() |
|
160
|
|
|
has_pubchem = any((isinstance(what, PubchemSearch) for what in self.what)) |
|
|
|
|
|
|
161
|
|
|
has_chembl = any((isinstance(what, ChemblSearch) for what in self.what)) |
|
162
|
|
|
# find the compounds first so the user knows what's missing before proceeding |
|
163
|
|
|
SearcherUtils.dl(inchikeys, pubchem=has_pubchem, chembl=has_chembl) |
|
164
|
|
|
for what in self.what: |
|
165
|
|
|
output_path = self.output_paths[what.key] |
|
166
|
|
|
df = what.find_to_df(inchikeys) |
|
|
|
|
|
|
167
|
|
|
# TODO keep any other columns in input_df |
|
|
|
|
|
|
168
|
|
|
df.to_csv(output_path) |
|
169
|
|
|
params = {k: str(v) for k, v in what.get_params().items() if k not in {"key", "api"}} |
|
170
|
|
|
metadata = NestedDotDict(dict(key=what.key, search=what.search_class, params=params)) |
|
171
|
|
|
metadata.write_json(output_path.with_suffix(".json")) |
|
172
|
|
|
logger.notice(f"Wrote {what.key} to {output_path}") |
|
173
|
|
|
return self |
|
174
|
|
|
|
|
175
|
|
|
def _output_path_of(self, what: Search, to: Optional[Path]) -> Path: |
|
|
|
|
|
|
176
|
|
|
if to is None: |
|
|
|
|
|
|
177
|
|
|
return self._default_path_of(what) |
|
178
|
|
|
elif str(to).startswith("."): |
|
179
|
|
|
return self._default_path_of(what).with_suffix(str(to)) |
|
180
|
|
|
else: |
|
181
|
|
|
return to |
|
182
|
|
|
|
|
183
|
|
|
def _default_path_of(self, what: Search) -> Path: |
|
184
|
|
|
parent = self.input_path.parent / (self.input_path.stem + "-output") |
|
185
|
|
|
parent.mkdir(exist_ok=True) |
|
186
|
|
|
child = what.key + ".csv" |
|
187
|
|
|
node = PathTools.sanitize_path_node(child) |
|
188
|
|
|
if (parent / node).resolve() != (parent / child).resolve(): |
|
189
|
|
|
logger.debug(f"Path {child} sanitized to {node}") |
|
190
|
|
|
return parent / node |
|
191
|
|
|
|
|
192
|
|
|
|
|
193
|
|
|
__all__ = ["Searcher", "IdMatchFrame", "SearcherUtils"] |
|
194
|
|
|
|
dmyersturnbull /
mandos
Loading history...