|
1
|
|
|
import enum |
|
|
|
|
|
|
2
|
|
|
import logging |
|
3
|
|
|
import abc |
|
4
|
|
|
from dataclasses import dataclass |
|
5
|
|
|
from typing import Sequence, Set, Optional |
|
6
|
|
|
|
|
7
|
|
|
from pocketutils.core.dot_dict import NestedDotDict |
|
|
|
|
|
|
8
|
|
|
|
|
9
|
|
|
from mandos.model.chembl_api import ChemblApi |
|
10
|
|
|
from mandos.model.chembl_support import ChemblCompound |
|
11
|
|
|
from mandos.model.chembl_support.chembl_target_graphs import ChemblTargetGraph |
|
12
|
|
|
from mandos.model.taxonomy import Taxonomy |
|
13
|
|
|
from mandos.search.chembl._protein_search import ProteinHit, ProteinSearch, H |
|
14
|
|
|
|
|
15
|
|
|
logger = logging.getLogger("mandos") |
|
16
|
|
|
|
|
17
|
|
|
|
|
18
|
|
|
class AssayType(enum.Enum): |
|
|
|
|
|
|
19
|
|
|
binding = enum.auto() |
|
20
|
|
|
functional = enum.auto() |
|
21
|
|
|
adme = enum.auto() |
|
22
|
|
|
physicochemical = enum.auto() |
|
23
|
|
|
|
|
24
|
|
|
@property |
|
25
|
|
|
def character(self) -> str: |
|
|
|
|
|
|
26
|
|
|
return { |
|
27
|
|
|
AssayType.binding: "B", |
|
28
|
|
|
AssayType.functional: "F", |
|
29
|
|
|
AssayType.adme: "A", |
|
30
|
|
|
AssayType.physicochemical: "P", |
|
31
|
|
|
}[self] |
|
32
|
|
|
|
|
33
|
|
|
|
|
34
|
|
|
@dataclass(frozen=True, order=True, repr=True) |
|
35
|
|
|
class _ActivityHit(ProteinHit): |
|
36
|
|
|
taxon_id: int |
|
37
|
|
|
taxon_name: str |
|
38
|
|
|
src_id: str |
|
39
|
|
|
|
|
40
|
|
|
|
|
41
|
|
|
class _ActivitySearch(ProteinSearch[H], metaclass=abc.ABCMeta): |
|
42
|
|
|
""" |
|
43
|
|
|
Search for ``activity``. |
|
44
|
|
|
""" |
|
45
|
|
|
|
|
46
|
|
|
def __init__( |
|
|
|
|
|
|
47
|
|
|
self, |
|
|
|
|
|
|
48
|
|
|
key: str, |
|
|
|
|
|
|
49
|
|
|
api: ChemblApi, |
|
|
|
|
|
|
50
|
|
|
taxa: Sequence[Taxonomy], |
|
|
|
|
|
|
51
|
|
|
traversal_strategy: str, |
|
|
|
|
|
|
52
|
|
|
allowed_target_types: Set[str], |
|
|
|
|
|
|
53
|
|
|
min_confidence_score: Optional[int], |
|
|
|
|
|
|
54
|
|
|
allowed_relations: Set[str], |
|
|
|
|
|
|
55
|
|
|
min_pchembl: Optional[float], |
|
|
|
|
|
|
56
|
|
|
banned_flags: Set[str], |
|
|
|
|
|
|
57
|
|
|
): |
|
58
|
|
|
super().__init__(key, api, taxa, traversal_strategy, allowed_target_types) |
|
59
|
|
|
self.min_confidence_score = min_confidence_score |
|
60
|
|
|
self.allowed_relations = allowed_relations |
|
61
|
|
|
self.min_pchembl = min_pchembl |
|
62
|
|
|
self.banned_flags = banned_flags |
|
63
|
|
|
|
|
64
|
|
|
@classmethod |
|
65
|
|
|
def allowed_assay_types(cls) -> Set[str]: |
|
|
|
|
|
|
66
|
|
|
raise NotImplementedError() |
|
67
|
|
|
|
|
68
|
|
|
def query(self, parent_form: ChemblCompound) -> Sequence[NestedDotDict]: |
|
|
|
|
|
|
69
|
|
|
|
|
70
|
|
|
filters = dict( |
|
71
|
|
|
parent_molecule_chembl_id=parent_form.chid, |
|
72
|
|
|
assay_type__iregex=self._set_to_regex(self.allowed_assay_types()), |
|
73
|
|
|
standard_relation__iregex=self._set_to_regex(self.allowed_relations), |
|
74
|
|
|
pchembl_value__isnull=None if self.min_pchembl is None else False, |
|
75
|
|
|
target_organism__isnull=None if len(self.taxa) == 0 else False, |
|
76
|
|
|
) |
|
77
|
|
|
# I'd rather not figure out how the API interprets None, so remove them |
|
78
|
|
|
filters = {k: v for k, v in filters.items() if v is not None} |
|
79
|
|
|
return list(self.api.activity.filter(**filters)) |
|
80
|
|
|
|
|
81
|
|
|
def should_include( |
|
|
|
|
|
|
82
|
|
|
self, lookup: str, compound: ChemblCompound, data: NestedDotDict, target: ChemblTargetGraph |
|
|
|
|
|
|
83
|
|
|
) -> bool: |
|
84
|
|
|
if ( |
|
85
|
|
|
( |
|
|
|
|
|
|
86
|
|
|
data.get_as("data_validity_comment", lambda s: s.lower()) |
|
87
|
|
|
in {s.lower() for s in self.banned_flags} |
|
88
|
|
|
) |
|
89
|
|
|
or (data.req_as("standard_relation", str) not in self.allowed_relations) |
|
|
|
|
|
|
90
|
|
|
or (data.req_as("assay_type", str) not in self.allowed_assay_types()) |
|
|
|
|
|
|
91
|
|
|
or (len(self.taxa) > 0 and not self.is_in_taxa(data.get_as("target_tax_id", int))) |
|
|
|
|
|
|
92
|
|
|
or (self.min_pchembl is not None and data.get("pchembl_value") is None) |
|
|
|
|
|
|
93
|
|
|
or self.min_pchembl is not None |
|
|
|
|
|
|
94
|
|
|
and data.req_as("pchembl_value", float) < self.min_pchembl |
|
|
|
|
|
|
95
|
|
|
): |
|
96
|
|
|
return False |
|
97
|
|
|
if data.get("data_validity_comment") is not None: |
|
98
|
|
|
logger.warning( |
|
|
|
|
|
|
99
|
|
|
f"Activity annotation for {lookup} has flag '{data.get('data_validity_comment')} (ok)" |
|
|
|
|
|
|
100
|
|
|
) |
|
101
|
|
|
# The `target_organism` doesn't always match the `assay_organism` |
|
102
|
|
|
# Ex: see assay CHEMBL823141 / document CHEMBL1135642 for homo sapiens in xenopus laevis |
|
103
|
|
|
# However, it's often something like yeast expressing a human / mouse / etc receptor |
|
104
|
|
|
# So there's no need to filter by it |
|
105
|
|
|
assay = self.api.assay.get(data.req_as("assay_chembl_id", str)) |
|
106
|
|
|
if target.type.name.lower() not in {s.lower() for s in self.allowed_target_types}: |
|
107
|
|
|
logger.warning(f"Excluding {target} with type {target.type}") |
|
|
|
|
|
|
108
|
|
|
return False |
|
109
|
|
|
confidence_score = assay.get("confidence_score") |
|
110
|
|
|
if self.min_confidence_score is not None: |
|
111
|
|
|
if confidence_score is None or confidence_score < self.min_confidence_score: |
|
112
|
|
|
return False |
|
113
|
|
|
# Some of these are non-protein types |
|
114
|
|
|
# And if it's unknown, we don't know what to do with it |
|
115
|
|
|
return True |
|
116
|
|
|
|
|
117
|
|
|
def _extract(self, lookup: str, compound: ChemblCompound, data: NestedDotDict) -> NestedDotDict: |
|
118
|
|
|
# we know these exist from the query |
|
119
|
|
|
organism = data.req_as("target_organism", str) |
|
120
|
|
|
tax_id = data.req_as("target_tax_id", int) |
|
121
|
|
|
if len(self.taxa) == 0: |
|
122
|
|
|
tax_id, tax_name = tax_id, organism |
|
|
|
|
|
|
123
|
|
|
else: |
|
124
|
|
|
taxes = {tax.req(tax_id) for tax in self.taxa if tax.contains(tax_id)} |
|
125
|
|
|
tax = next(iter(taxes)) |
|
126
|
|
|
if len(taxes) > 1: |
|
127
|
|
|
logger.warning(f"Multiple matches for taxon {tax_id}: {taxes}; using {tax}") |
|
|
|
|
|
|
128
|
|
|
if organism != tax.name: |
|
129
|
|
|
logger.warning(f"Target organism {organism} is not {tax.name}") |
|
|
|
|
|
|
130
|
|
|
tax_id = tax.id |
|
131
|
|
|
tax_name = tax.name |
|
132
|
|
|
return NestedDotDict( |
|
133
|
|
|
dict( |
|
134
|
|
|
record_id=data.req_as("activity_id", str), |
|
135
|
|
|
origin_inchikey=lookup, |
|
136
|
|
|
matched_inchikey=compound.inchikey, |
|
137
|
|
|
compound_id=compound.chid, |
|
138
|
|
|
compound_name=compound.name, |
|
139
|
|
|
taxon_id=tax_id, |
|
140
|
|
|
taxon_name=tax_name, |
|
141
|
|
|
pchembl=data.req_as("pchembl_value", float), |
|
142
|
|
|
std_type=data.req_as("standard_type", str), |
|
143
|
|
|
src_id=data.req_as("src_id", str), |
|
144
|
|
|
exact_target_id=data.req_as("target_chembl_id", str), |
|
145
|
|
|
tissue=data.get_as("tissue", str), |
|
146
|
|
|
cell_type=data.get_as("cell_type", str), |
|
147
|
|
|
subcellular_region=data.get("subcellular_region", str), |
|
148
|
|
|
) |
|
149
|
|
|
) |
|
150
|
|
|
|
|
151
|
|
|
assay_type: AssayType |
|
152
|
|
|
tissue: Optional[str] |
|
153
|
|
|
cell_type: Optional[str] |
|
154
|
|
|
subcellular_region: Optional[str] |
|
155
|
|
|
|
|
156
|
|
|
|
|
157
|
|
|
__all__ = ["_ActivitySearch", "AssayType", "_ActivityHit"] |
|
158
|
|
|
|
dmyersturnbull /
mandos
Loading history...