Skip to content

Commit

Permalink
keep working
Browse files Browse the repository at this point in the history
  • Loading branch information
@SalvadorBrandolin
SalvadorBrandolin committed Sep 9, 2024
1 parent 678160e commit 5f3719e
Show file tree
Hide file tree
Showing 3 changed files with 102 additions and 283 deletions.
55 changes: 55 additions & 0 deletions ugropy/fragmentation_models/fragmentation_gibbs_model.py
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,55 @@
"""GibbsModel module."""

from typing import Union

import pandas as pd

from ugropy.fragmentation_models.fragmentation_model import FragmentationModel


class GibbsModel(FragmentationModel):
"""GibbsModel it's a fragmentation model dedicated to Gibbs excess models.
unifac, psrk, dortmund are instances of this class.
Parameters
----------
subgroups : pd.DataFrame
Model's subgroups. Index: 'group' (subgroups names). Mandatory columns:
'smarts' (SMARTS representations of the group to detect its precense in
the molecule), 'molecular_weight' (molecular weight of the subgroups
used to check that the result is correct)
subgroups_info : Union[pd.DataFrame, None], optional
Information of the model's subgroups (R, Q, subgroup_number,
main_group), by default None
Attributes
----------
subgroups : pd.DataFrame
Model's subgroups. Index: 'group' (subgroups names). Mandatory columns:
'smarts' (SMARTS representations of the group to detect its precense in
the molecule), 'molecular_weight' (molecular weight of the subgroups
used to check that the result is correct)
detection_mols : dict
Dictionary cotaining all the rdkit Mol object from the detection_smarts
subgroups column
subgroups_info : pd.DataFrame
Information of the model's subgroups. Columns: R, Q, subgroup_number,
main_group. Index: 'group' (subgroups names)
"""

def __init__(
self,
subgroups: pd.DataFrame,
subgroups_info: Union[pd.DataFrame, None] = None,
) -> None:
super().__init__(subgroups)

# subgroups info
if subgroups_info is None:
self.subgroups_info = pd.DataFrame(
[],
columns=["group", "subgroup_number", "main_group", "R", "Q"],
).set_index("group")
else:
self.subgroups_info = subgroups_info
219 changes: 47 additions & 172 deletions ugropy/fragmentation_models/fragmentation_model.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,206 +3,81 @@
All ugropy models (joback, unifac, psrk) are instances of the
FragmentationModule class.
"""

import json
from typing import List, Union

import numpy as np
from abc import ABC, abstractmethod

import pandas as pd

from rdkit import Chem


class FragmentationModel:
class FragmentationModel(ABC):
"""FragmentationModel class.
All ugropy supported models are an instance of this class. This class can
be used by the user to create their own FragmentationModels.
All ugropy supported models are an instance of this class. This class must
be inherited to create a new type of FragmentationModel.
Parameters
----------
subgroups : pd.DataFrame
Model's subgroups. Index: 'group' (groups names). Mandatory columns:
'detection_smarts' (SMARTS representations of the group to detect its
precense in the molecule), 'smarts' (true SMARTS of the group witouht
additional atom detections), 'contribute' (dictionary as a string with
the group contribution), 'composed' (y or n if it is or is not a
composed structure), 'molecular_weight' (molecular weight of the
subgroups).
split_detection_smarts : List[str], optional
List of subgroups that have different SMARTS representations. by
default []
problematic_structures : Union[pd.DataFrame, None], optional
Model's problematic/ambiguous structures. Index: 'smarts' (SMARTS of
the problematic structure). Mandatory columns: 'contribute' (dictionary
as a string with the structure contribution), by default None
hideouts : Union[pd.DataFrame, None], optional
Hideouts for each group. Index: 'group' (Group of the model that can be
hiden). Mandatory columns: 'hideout' (other subgroups to find the hiden
subgroup), by defautl None
Model's subgroups. Index: 'group' (subgroups names). Mandatory columns:
'smarts' (SMARTS representations of the group to detect its precense in
the molecule), 'molecular_weight' (molecular weight of the subgroups
used to check that the result is correct).
Attributes
----------
subgroups : pd.DataFrame
Model's subgroups. Index: 'group' (groups names). Mandatory columns:
'detection_smarts' (SMARTS representations of the group to detect its
precense in the molecule), 'smarts' (true SMARTS of the group witouht
additional atom detections. If a value is missing uses the
corresponding detection_smarts), 'contribute' (dictionary as a string
with the group contribution), 'composed' (y or n if it is or is not a
composed structure), 'molecular_weight' (molecular weight of the
subgroups).
split_detection_smarts : List[str]
List of subgroups that have different SMARTS representations.
problematic_structures : pd.Dataframe
Model's problematic/ambiguous structures. Index: 'smarts' (SMARTS of
the problematic structure). Mandatory columns: 'contribute' (dictionary
as a string with the structure contribution)
hideouts : pd.DataFrame
Hideouts for each group. Index: 'group' (Group of the model that can be
hiden). Mandatory columns: 'hideout' (other subgroups to find the hiden
subgroup)
Model's subgroups. Index: 'group' (subgroups names). Mandatory columns:
'smarts' (SMARTS representations of the group to detect its precense in
the molecule), 'molecular_weight' (molecular weight of the subgroups
used to check that the result is correct).
detection_mols : dict
Dictionary cotaining all the rdkit Mol object from the detection_smarts
subgroups column.
fit_mols : dict
Dictionary cotaining all the rdkit Mol object from the smarts subgroups
column.
contribution_matrix : pd.Dataframe
Contribution matrix of the model built from the subgroups contribute.
"""

def __init__(
self,
subgroups: pd.DataFrame,
split_detection_smarts: List[str] = [],
problematic_structures: Union[pd.DataFrame, None] = None,
hideouts: Union[pd.DataFrame, None] = None,
) -> None:
def __init__(self, subgroups: pd.DataFrame) -> None:
self.subgroups = subgroups
self.split_detection_smarts = split_detection_smarts

# =====================================================================
# Empty problematics template
# =====================================================================
if problematic_structures is None:
self.problematic_structures = pd.DataFrame(
[], columns=["smarts", "contribute"]
).set_index("smarts")
else:
self.problematic_structures = problematic_structures

# =====================================================================
# Hideouts
# =====================================================================
if hideouts is None:
self.hideouts = pd.DataFrame(
[], columns=["group", "hideout"]
).set_index("group")
else:
self.hideouts = hideouts

# =====================================================================
# Contribution matrix build
# =====================================================================
self.contribution_matrix = self._build_contrib_matrix()

# =====================================================================
# Instantiate all de mol object from their smarts
# =====================================================================
self.detection_mols = self._instantiate_detection_mol()
self.fit_mols = self._instantiate_fit_mols()

def _build_contrib_matrix(self) -> pd.DataFrame:
"""Build the contribution matrix of the model.
Returns
-------
pd.DataFrame
The contribution matrix of the model built from the contribute
column of the subgroups DataFrame.
Raises
------
ValueError
Bad contribution parsing of a group
TypeError
Bad contribution parsing of a group
"""
index = self.subgroups.index.to_numpy()
matrix = np.zeros((len(index), len(index)), dtype=int)

# Build the matrix
dfm = pd.DataFrame(matrix, index=index, columns=index).rename_axis(
"group"
)

# Instantiate all de mol object from their smarts representation
detection_mols = {}

# Fill the matrix
for group in index:
str_contribution = self.subgroups.loc[group, "contribute"]
for group, row in self.subgroups.iterrows():
detection_mols[group] = Chem.MolFromSmarts(row["smarts"])

try:
contribution = json.loads(str_contribution)
except json.JSONDecodeError:
raise ValueError(
f"Bad contribute parsing of the group: {group}"
)
except TypeError:
raise TypeError(
f"Bad contribution parsing of the group: {group}."
)

def detect_groups(self, molecule: Chem.Mol) -> pd.DataFrame:
"""Detect all the groups in the molecule.
for k in contribution.keys():
dfm.loc[group, k] = contribution[k]
Return a dictionary with the detected groups as keys and a tuple of
tuples containing the molecule's atoms that participate in the group
occurrences.
return dfm

def _instantiate_detection_mol(self) -> dict:
"""Instantiate all the rdkit Mol object from the detection_smarts.
Returns
-------
dict
Mol objects.
"""
mols = {}

for group in self.subgroups.index:
if group not in self.split_detection_smarts:
mols[group] = [
Chem.MolFromSmarts(
self.subgroups.loc[group, "detection_smarts"]
)
]
else:
smarts = self.subgroups.loc[group, "detection_smarts"].split(
","
)

mol_smarts = []
for sms in smarts:
mol_smarts.append(Chem.MolFromSmarts(sms))

mols[group] = mol_smarts
return mols

def _instantiate_fit_mols(self) -> dict:
"""Instantiate all the rdkit Mol object from the smarts.
Parameters
----------
mol : Chem.Mol
Molecule to detect the groups.
Returns
-------
dict
Mol object to fit the subgroups in the molecule's atoms.
Detected groups in the molecule.
"""
mols = {}

for group in self.subgroups.index:
smarts = self.subgroups.loc[group, "smarts"]

if isinstance(smarts, str):
mols[group] = [Chem.MolFromSmarts(smarts)]
else:
mols[group] = self.detection_mols[group]

return mols
detected_groups = {}
for group, mol in self.detection_mols.items():
matches = molecule.GetSubstructMatches(mol)

if matches:
detected_groups[group] = matches

return detected_groups

@abstractmethod
def set_fragmentation_result(
self,
molecule: Chem.Mol,
subgroups_occurrences: dict,
subgroups_atoms_indexes: dict,
) -> "FragmentationResult":

raise NotImplementedError("Abstract Method not implemented.")
Loading

0 comments on commit 5f3719e

Please sign in to comment.