Spaces:
Running
Running
New tools and filters for cheminfo
Browse files* Update Cheminformatrics use Cases
Add _cheminfo_tools.py with lipinksi filter , View Mol Image , View mol filter with smarts and smiles and highlights are done .
* update new filters and chembl webapi
update new filters and chembl webapi
veber, pains, muegge, brenk_aggregator_filter, egan , ghose , new qsar2.py code with matplotlib plots.
* update tools
update on chembl uniprot based search
* update the code
Delete the old files and folder
Put in example \ Cheminformatics folders
Chembl web service client with example
Plots with plot qsar and plot qsar2 with confidence intervals
* Update new code with new workspace
New workspace created deleted ex1 and ex2 .
Deleted the ecfp and maccs model .pkl file
- examples/.crdt/Image table.lynxkite.json.crdt +0 -0
- examples/.crdt/requirements.txt.crdt +0 -0
- examples/Cheminformatics/chem_utils.py +263 -0
- examples/Cheminformatics/chembl_api_uses.lynxkite.json +0 -0
- examples/Cheminformatics/chembl_tools.py +206 -0
- examples/Cheminformatics/cheminfo_tools.py +610 -0
- examples/Cheminformatics/qsar_example.lynxkite.json +0 -0
- examples/draw_molecules.py +0 -29
- examples/requirements.txt +3 -0
examples/.crdt/Image table.lynxkite.json.crdt
ADDED
Binary file (31.8 kB). View file
|
|
examples/.crdt/requirements.txt.crdt
ADDED
Binary file (251 Bytes). View file
|
|
examples/Cheminformatics/chem_utils.py
ADDED
@@ -0,0 +1,263 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import base64
|
2 |
+
import io
|
3 |
+
import sys
|
4 |
+
from io import StringIO
|
5 |
+
from operator import itemgetter
|
6 |
+
from typing import List
|
7 |
+
from typing import Tuple
|
8 |
+
import itertools
|
9 |
+
import matplotlib.pyplot as plt
|
10 |
+
import numpy as np
|
11 |
+
import seaborn as sns
|
12 |
+
from rdkit import Chem, DataStructs, RDLogger
|
13 |
+
from rdkit.Chem.Draw import rdMolDraw2D
|
14 |
+
from rdkit.Chem.rdchem import Mol
|
15 |
+
from rdkit.ML.Cluster import Butina
|
16 |
+
from rdkit.rdBase import BlockLogs
|
17 |
+
|
18 |
+
import pandas as pd
|
19 |
+
from rdkit.Chem.rdMMPA import FragmentMol
|
20 |
+
from rdkit.Chem.rdRGroupDecomposition import RGroupDecompose
|
21 |
+
|
22 |
+
|
23 |
+
def smi2mol_with_errors(smi: str) -> Tuple[Mol, str]:
|
24 |
+
"""Parse SMILES and return any associated errors or warnings
|
25 |
+
|
26 |
+
:param smi: input SMILES
|
27 |
+
:return: tuple of RDKit molecule, warning or error
|
28 |
+
"""
|
29 |
+
sio = sys.stderr = StringIO()
|
30 |
+
mol = Chem.MolFromSmiles(smi)
|
31 |
+
err = sio.getvalue()
|
32 |
+
sio = sys.stderr = StringIO()
|
33 |
+
sys.stderr = sys.__stderr__
|
34 |
+
return mol, err
|
35 |
+
|
36 |
+
|
37 |
+
def count_fragments(mol: Mol) -> int:
|
38 |
+
"""Count the number of fragments in a molecule
|
39 |
+
|
40 |
+
:param mol: RDKit molecule
|
41 |
+
:return: number of fragments
|
42 |
+
"""
|
43 |
+
return len(Chem.GetMolFrags(mol, asMols=True))
|
44 |
+
|
45 |
+
|
46 |
+
def get_largest_fragment(mol: Mol) -> Mol:
|
47 |
+
"""Return the fragment with the largest number of atoms
|
48 |
+
|
49 |
+
:param mol: RDKit molecule
|
50 |
+
:return: RDKit molecule with the largest number of atoms
|
51 |
+
"""
|
52 |
+
frag_list = list(Chem.GetMolFrags(mol, asMols=True))
|
53 |
+
frag_mw_list = [(x.GetNumAtoms(), x) for x in frag_list]
|
54 |
+
frag_mw_list.sort(key=itemgetter(0), reverse=True)
|
55 |
+
return frag_mw_list[0][1]
|
56 |
+
|
57 |
+
|
58 |
+
# ----------- Clustering
|
59 |
+
# https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GroupShuffleSplit.html
|
60 |
+
def taylor_butina_clustering(
|
61 |
+
fp_list: List[DataStructs.ExplicitBitVect], cutoff: float = 0.65
|
62 |
+
) -> List[int]:
|
63 |
+
"""Cluster a set of fingerprints using the RDKit Taylor-Butina implementation
|
64 |
+
|
65 |
+
:param fp_list: a list of fingerprints
|
66 |
+
:param cutoff: distance cutoff (1 - Tanimoto similarity)
|
67 |
+
:return: a list of cluster ids
|
68 |
+
"""
|
69 |
+
dists = []
|
70 |
+
nfps = len(fp_list)
|
71 |
+
for i in range(1, nfps):
|
72 |
+
sims = DataStructs.BulkTanimotoSimilarity(fp_list[i], fp_list[:i])
|
73 |
+
dists.extend([1 - x for x in sims])
|
74 |
+
cluster_res = Butina.ClusterData(dists, nfps, cutoff, isDistData=True)
|
75 |
+
cluster_id_list = np.zeros(nfps, dtype=int)
|
76 |
+
for cluster_num, cluster in enumerate(cluster_res):
|
77 |
+
for member in cluster:
|
78 |
+
cluster_id_list[member] = cluster_num
|
79 |
+
return cluster_id_list.tolist()
|
80 |
+
|
81 |
+
|
82 |
+
# ----------- Atom tagging
|
83 |
+
def label_atoms(mol: Mol, labels: List[str]) -> Mol:
|
84 |
+
"""Label atoms when depicting a molecule
|
85 |
+
|
86 |
+
:param mol: input molecule
|
87 |
+
:param labels: labels, one for each atom
|
88 |
+
:return: molecule with labels
|
89 |
+
"""
|
90 |
+
[atm.SetProp("atomNote", "") for atm in mol.GetAtoms()]
|
91 |
+
for atm in mol.GetAtoms():
|
92 |
+
idx = atm.GetIdx()
|
93 |
+
mol.GetAtomWithIdx(idx).SetProp("atomNote", f"{labels[idx]}")
|
94 |
+
return mol
|
95 |
+
|
96 |
+
|
97 |
+
def tag_atoms(mol: Mol, atoms_to_tag: List[int], tag: str = "x") -> Mol:
|
98 |
+
"""Tag atoms with a specified string
|
99 |
+
|
100 |
+
:param mol: input molecule
|
101 |
+
:param atoms_to_tag: indices of atoms to tag
|
102 |
+
:param tag: string to use for the tags
|
103 |
+
:return: molecule with atoms tagged
|
104 |
+
"""
|
105 |
+
[atm.SetProp("atomNote", "") for atm in mol.GetAtoms()]
|
106 |
+
[mol.GetAtomWithIdx(idx).SetProp("atomNote", tag) for idx in atoms_to_tag]
|
107 |
+
return mol
|
108 |
+
|
109 |
+
|
110 |
+
# ----------- Logging
|
111 |
+
def rd_shut_the_hell_up() -> None:
|
112 |
+
"""Make the RDKit be a bit more quiet
|
113 |
+
|
114 |
+
:return: None
|
115 |
+
"""
|
116 |
+
lg = RDLogger.logger()
|
117 |
+
lg.setLevel(RDLogger.CRITICAL)
|
118 |
+
|
119 |
+
|
120 |
+
def demo_block_logs() -> None:
|
121 |
+
"""An example of another way to turn off RDKit logging
|
122 |
+
|
123 |
+
:return: None
|
124 |
+
"""
|
125 |
+
block = BlockLogs()
|
126 |
+
# do stuff
|
127 |
+
del block
|
128 |
+
|
129 |
+
|
130 |
+
# ----------- Image generation
|
131 |
+
def boxplot_base64_image(dist: np.ndarray, x_lim: list[int] = [0, 10]) -> str:
|
132 |
+
"""
|
133 |
+
Plot a distribution as a seaborn boxplot and save the resulting image as a base64 image.
|
134 |
+
|
135 |
+
Parameters:
|
136 |
+
dist (np.ndarray): The distribution data to plot.
|
137 |
+
x_lim (list[int]): The x-axis limits for the boxplot.
|
138 |
+
|
139 |
+
Returns:
|
140 |
+
str: The base64 encoded image string.
|
141 |
+
"""
|
142 |
+
sns.set(rc={"figure.figsize": (3, 1)})
|
143 |
+
sns.set_style("whitegrid")
|
144 |
+
ax = sns.boxplot(x=dist)
|
145 |
+
ax.set_xlim(x_lim[0], x_lim[1])
|
146 |
+
s = io.BytesIO()
|
147 |
+
plt.savefig(s, format="png", bbox_inches="tight")
|
148 |
+
plt.close()
|
149 |
+
s = base64.b64encode(s.getvalue()).decode("utf-8").replace("\n", "")
|
150 |
+
return '<img align="left" src="data:image/png;base64,%s">' % s
|
151 |
+
|
152 |
+
|
153 |
+
def mol_to_base64_image(mol: Chem.Mol) -> str:
|
154 |
+
"""
|
155 |
+
Convert an RDKit molecule to a base64 encoded image string.
|
156 |
+
|
157 |
+
Parameters:
|
158 |
+
mol (Chem.Mol): The RDKit molecule to convert.
|
159 |
+
|
160 |
+
Returns:
|
161 |
+
str: The base64 encoded image string.
|
162 |
+
"""
|
163 |
+
drawer = rdMolDraw2D.MolDraw2DCairo(300, 150)
|
164 |
+
drawer.DrawMolecule(mol)
|
165 |
+
drawer.FinishDrawing()
|
166 |
+
text = drawer.GetDrawingText()
|
167 |
+
im_text64 = base64.b64encode(text).decode("utf8")
|
168 |
+
img_str = f"<img src='data:image/png;base64, {im_text64}'/>"
|
169 |
+
return img_str
|
170 |
+
|
171 |
+
|
172 |
+
def cleanup_fragment(mol: Mol) -> Tuple[Mol, int]:
|
173 |
+
"""
|
174 |
+
Replace atom map numbers with Hydrogens
|
175 |
+
:param mol: input molecule
|
176 |
+
:return: modified molecule, number of R-groups
|
177 |
+
"""
|
178 |
+
rgroup_count = 0
|
179 |
+
for atm in mol.GetAtoms():
|
180 |
+
atm.SetAtomMapNum(0)
|
181 |
+
if atm.GetAtomicNum() == 0:
|
182 |
+
rgroup_count += 1
|
183 |
+
atm.SetAtomicNum(1)
|
184 |
+
mol = Chem.RemoveAllHs(mol)
|
185 |
+
return mol, rgroup_count
|
186 |
+
|
187 |
+
|
188 |
+
def generate_fragments(mol: Mol) -> pd.DataFrame:
|
189 |
+
"""
|
190 |
+
Generate fragments using the RDKit
|
191 |
+
:param mol: RDKit molecule
|
192 |
+
:return: a Pandas dataframe with Scaffold SMILES, Number of Atoms, Number of R-Groups
|
193 |
+
"""
|
194 |
+
# Generate molecule fragments
|
195 |
+
frag_list = FragmentMol(mol)
|
196 |
+
# Flatten the output into a single list
|
197 |
+
flat_frag_list = [x for x in itertools.chain(*frag_list) if x]
|
198 |
+
# The output of Fragment mol is contained in single molecules. Extract the largest fragment from each molecule
|
199 |
+
flat_frag_list = [get_largest_fragment(x) for x in flat_frag_list]
|
200 |
+
# Keep fragments where the number of atoms in the fragment is at least 2/3 of the number fragments in
|
201 |
+
# input molecule
|
202 |
+
num_mol_atoms = mol.GetNumAtoms()
|
203 |
+
flat_frag_list = [x for x in flat_frag_list if x.GetNumAtoms() / num_mol_atoms > 0.67]
|
204 |
+
# remove atom map numbers from the fragments
|
205 |
+
flat_frag_list = [cleanup_fragment(x) for x in flat_frag_list]
|
206 |
+
# Convert fragments to SMILES
|
207 |
+
frag_smiles_list = [[Chem.MolToSmiles(x), x.GetNumAtoms(), y] for (x, y) in flat_frag_list]
|
208 |
+
# Add the input molecule to the fragment list
|
209 |
+
frag_smiles_list.append([Chem.MolToSmiles(mol), mol.GetNumAtoms(), 1])
|
210 |
+
# Put the results into a Pandas dataframe
|
211 |
+
frag_df = pd.DataFrame(frag_smiles_list, columns=["Scaffold", "NumAtoms", "NumRgroupgs"])
|
212 |
+
# Remove duplicate fragments
|
213 |
+
frag_df = frag_df.drop_duplicates("Scaffold")
|
214 |
+
return frag_df
|
215 |
+
|
216 |
+
|
217 |
+
def find_scaffolds(df_in: pd.DataFrame) -> Tuple[pd.DataFrame, pd.DataFrame]:
|
218 |
+
"""
|
219 |
+
Generate scaffolds for a set of molecules
|
220 |
+
:param df_in: Pandas dataframe with [SMILES, Name, RDKit molecule] columns
|
221 |
+
:return: dataframe with molecules and scaffolds, dataframe with unique scaffolds
|
222 |
+
"""
|
223 |
+
# Loop over molecules and generate fragments, fragments for each molecule are returned as a Pandas dataframe
|
224 |
+
df_list = []
|
225 |
+
for smiles, name, mol in df_in[["SMILES", "Name", "mol"]].values:
|
226 |
+
tmp_df = generate_fragments(mol).copy()
|
227 |
+
tmp_df["Name"] = name
|
228 |
+
tmp_df["SMILES"] = smiles
|
229 |
+
df_list.append(tmp_df)
|
230 |
+
# Combine the list of dataframes into a single dataframe
|
231 |
+
mol_df = pd.concat(df_list)
|
232 |
+
# Collect scaffolds
|
233 |
+
scaffold_list = []
|
234 |
+
for k, v in mol_df.groupby("Scaffold"):
|
235 |
+
scaffold_list.append([k, len(v.Name.unique()), v.NumAtoms.values[0]])
|
236 |
+
scaffold_df = pd.DataFrame(scaffold_list, columns=["Scaffold", "Count", "NumAtoms"])
|
237 |
+
# Any fragment that occurs more times than the number of fragments can't be a scaffold
|
238 |
+
num_df_rows = len(df_in) # noqa: F841
|
239 |
+
scaffold_df = scaffold_df.query(f"Count <= {num_df_rows}")
|
240 |
+
# Sort scaffolds by frequency
|
241 |
+
scaffold_df = scaffold_df.sort_values(["Count", "NumAtoms"], ascending=[False, False])
|
242 |
+
return mol_df, scaffold_df
|
243 |
+
|
244 |
+
|
245 |
+
def get_molecules_with_scaffold(
|
246 |
+
scaffold: str, mol_df: pd.DataFrame, activity_df: pd.DataFrame
|
247 |
+
) -> Tuple[List[str], pd.DataFrame]:
|
248 |
+
"""
|
249 |
+
Associate molecules with scaffolds
|
250 |
+
:param scaffold: scaffold SMILES
|
251 |
+
:param mol_df: dataframe with molecules and scaffolds, returned by find_scaffolds()
|
252 |
+
:param activity_df: dataframe with [SMILES, Name, pIC50] columns
|
253 |
+
:return: list of core(s) with R-groups labeled, dataframe with [SMILES, Name, pIC50]
|
254 |
+
"""
|
255 |
+
match_df = mol_df.query("Scaffold == @scaffold")
|
256 |
+
merge_df = match_df.merge(activity_df, on=["SMILES", "Name"])
|
257 |
+
scaffold_mol = Chem.MolFromSmiles(scaffold)
|
258 |
+
rgroup_match, rgroup_miss = RGroupDecompose(scaffold_mol, merge_df.mol, asSmiles=True)
|
259 |
+
if len(rgroup_match):
|
260 |
+
rgroup_df = pd.DataFrame(rgroup_match)
|
261 |
+
return rgroup_df.Core.unique(), merge_df[["SMILES", "Name", "pIC50"]]
|
262 |
+
else:
|
263 |
+
return [], merge_df[["SMILES", "Name", "pIC50"]]
|
examples/Cheminformatics/chembl_api_uses.lynxkite.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
examples/Cheminformatics/chembl_tools.py
ADDED
@@ -0,0 +1,206 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from lynxkite.core.ops import op
|
2 |
+
import pandas as pd
|
3 |
+
from chembl_webresource_client.new_client import new_client
|
4 |
+
from rdkit import Chem
|
5 |
+
|
6 |
+
|
7 |
+
@op("LynxKite Graph Analytics", "chembl sim search")
|
8 |
+
def similarity_to_dataframe(*, smiles: str, cutoff: int = 70) -> pd.DataFrame:
|
9 |
+
"""
|
10 |
+
Run a ChEMBL similarity search and return the hits as a pandas DataFrame.
|
11 |
+
If the SMILES is invalid or an error occurs, prints a message and returns
|
12 |
+
an empty DataFrame with the expected columns.
|
13 |
+
|
14 |
+
Parameters
|
15 |
+
----------
|
16 |
+
smiles : str
|
17 |
+
The SMILES string to search on.
|
18 |
+
cutoff : int
|
19 |
+
The minimum Tanimoto similarity (0–100).
|
20 |
+
|
21 |
+
Returns
|
22 |
+
-------
|
23 |
+
pd.DataFrame
|
24 |
+
Columns: 'molecule_chembl_id', 'similarity'
|
25 |
+
"""
|
26 |
+
# Prepare empty frame to return on error
|
27 |
+
cols = ["molecule_chembl_id", "similarity"]
|
28 |
+
empty_df = pd.DataFrame(columns=cols)
|
29 |
+
|
30 |
+
# 1) Quick SMILES validation
|
31 |
+
if Chem.MolFromSmiles(smiles) is None:
|
32 |
+
print("Please input a correct SMILES string.")
|
33 |
+
return empty_df
|
34 |
+
|
35 |
+
try:
|
36 |
+
# 2) Do the ChEMBL API call
|
37 |
+
similarity = new_client.similarity
|
38 |
+
results = similarity.filter(smiles=smiles, similarity=cutoff).only(cols)
|
39 |
+
|
40 |
+
# 3) Build DataFrame
|
41 |
+
data = list(results)
|
42 |
+
df = pd.DataFrame.from_records(data, columns=cols)
|
43 |
+
|
44 |
+
# 4) Inform if no hits
|
45 |
+
if df.empty:
|
46 |
+
print("No hits found for that SMILES at the given cutoff.")
|
47 |
+
return df
|
48 |
+
|
49 |
+
except Exception as e:
|
50 |
+
# Catch network errors, unexpected API replies, etc.
|
51 |
+
print("An error occurred during the similarity search.")
|
52 |
+
print(" Details:", str(e))
|
53 |
+
return empty_df
|
54 |
+
|
55 |
+
|
56 |
+
@op("LynxKite Graph Analytics", "chembl structure")
|
57 |
+
def _chembl_structures(
|
58 |
+
df: pd.DataFrame, *, id_col: str = "molecule_chembl_id", timeout: int = 5
|
59 |
+
) -> pd.DataFrame:
|
60 |
+
"""
|
61 |
+
Given a DataFrame with a column of ChEMBL molecule IDs, append
|
62 |
+
canonical SMILES, standard InChI, and standard InChIKey.
|
63 |
+
|
64 |
+
Parameters
|
65 |
+
----------
|
66 |
+
df : pd.DataFrame
|
67 |
+
Input DataFrame; must contain `id_col`.
|
68 |
+
id_col : str
|
69 |
+
Name of the column in `df` that holds ChEMBL IDs (e.g. 'CHEMBL1234').
|
70 |
+
timeout : int
|
71 |
+
How many seconds to wait for the API (not currently used by chembl client,
|
72 |
+
but reserved for future enhancements or custom wrappers).
|
73 |
+
|
74 |
+
Returns
|
75 |
+
-------
|
76 |
+
pd.DataFrame
|
77 |
+
A new DataFrame with three additional columns:
|
78 |
+
- smiles
|
79 |
+
- standard_inchi
|
80 |
+
- standard_inchi_key
|
81 |
+
"""
|
82 |
+
# make a copy so we don’t modify in-place
|
83 |
+
out = df.copy()
|
84 |
+
# prepare new columns
|
85 |
+
out["smiles"] = None
|
86 |
+
out["standard_inchi"] = None
|
87 |
+
out["standard_inchi_key"] = None
|
88 |
+
|
89 |
+
mol_client = new_client.molecule
|
90 |
+
|
91 |
+
for idx, chembl_id in out[id_col].items():
|
92 |
+
try:
|
93 |
+
# query ChEMBL for this molecule
|
94 |
+
res = mol_client.filter(chembl_id=chembl_id).only(
|
95 |
+
["molecule_chembl_id", "molecule_structures"]
|
96 |
+
)
|
97 |
+
# filter() returns an iterable; grab first record if exists
|
98 |
+
rec = next(iter(res), None)
|
99 |
+
if rec and rec.get("molecule_structures"):
|
100 |
+
struct = rec["molecule_structures"]
|
101 |
+
out.at[idx, "smiles"] = struct.get("canonical_smiles")
|
102 |
+
out.at[idx, "standard_inchi"] = struct.get("standard_inchi")
|
103 |
+
out.at[idx, "standard_inchi_key"] = struct.get("standard_inchi_key")
|
104 |
+
else:
|
105 |
+
print(f"[Warning] No structure found for {chembl_id}")
|
106 |
+
except Exception as e:
|
107 |
+
print(f"[Error] Lookup failed for {chembl_id}: {e!s}")
|
108 |
+
|
109 |
+
return out
|
110 |
+
|
111 |
+
|
112 |
+
@op("LynxKite Graph Analytics", "get chembl drugs")
|
113 |
+
def fetch_chembl_drugs(
|
114 |
+
*, first_approval: int = 2000, development_phase: int = None
|
115 |
+
) -> pd.DataFrame:
|
116 |
+
"""
|
117 |
+
Fetch drugs from ChEMBL matching the given USAN stem, approval year,
|
118 |
+
and development phase, returning key fields as a DataFrame.
|
119 |
+
|
120 |
+
Parameters
|
121 |
+
----------
|
122 |
+
first_approval : int, optional
|
123 |
+
Only include drugs first approved in or after this year (default=1980).
|
124 |
+
development_phase : int, optional
|
125 |
+
Only include drugs in this development phase (e.g. 2, 3, 4).
|
126 |
+
If None, do not filter by phase.
|
127 |
+
usan_stem : str, optional
|
128 |
+
USAN stem to filter on (default="-azosin").
|
129 |
+
|
130 |
+
Returns
|
131 |
+
-------
|
132 |
+
pd.DataFrame
|
133 |
+
Columns:
|
134 |
+
- development_phase
|
135 |
+
- first_approval
|
136 |
+
- molecule_chembl_id
|
137 |
+
- synonyms
|
138 |
+
- usan_stem
|
139 |
+
- usan_stem_definition
|
140 |
+
- usan_year
|
141 |
+
|
142 |
+
If no results (or on error), returns an empty DataFrame with these columns.
|
143 |
+
"""
|
144 |
+
cols = [
|
145 |
+
"development_phase",
|
146 |
+
"first_approval",
|
147 |
+
"molecule_chembl_id",
|
148 |
+
"synonyms",
|
149 |
+
"usan_stem",
|
150 |
+
"usan_stem_definition",
|
151 |
+
"usan_year",
|
152 |
+
]
|
153 |
+
empty_df = pd.DataFrame(columns=cols)
|
154 |
+
|
155 |
+
# Validate inputs
|
156 |
+
if first_approval is not None and not isinstance(first_approval, int):
|
157 |
+
print("Error: first_approval must be an integer year.")
|
158 |
+
return empty_df
|
159 |
+
if development_phase is not None and not isinstance(development_phase, int):
|
160 |
+
print("Error: development_phase must be an integer.")
|
161 |
+
return empty_df
|
162 |
+
# if not isinstance(usan_stem, str):
|
163 |
+
# print("Error: usan_stem must be a string.")
|
164 |
+
# return empty_df
|
165 |
+
|
166 |
+
try:
|
167 |
+
drug = new_client.drug
|
168 |
+
|
169 |
+
# apply approval-year filter
|
170 |
+
if first_approval is not None:
|
171 |
+
drug = drug.filter(first_approval__gte=first_approval)
|
172 |
+
# apply development-phase filter
|
173 |
+
if development_phase is not None:
|
174 |
+
drug = drug.filter(development_phase=development_phase)
|
175 |
+
# apply USAN stem filter
|
176 |
+
# drug = drug.filter(usan_stem=usan_stem)
|
177 |
+
|
178 |
+
res = drug.only(cols)
|
179 |
+
df = pd.DataFrame(res, columns=cols)
|
180 |
+
|
181 |
+
if df.empty:
|
182 |
+
print("No drugs found for those filters.")
|
183 |
+
return df
|
184 |
+
|
185 |
+
except Exception as e:
|
186 |
+
print("An error occurred during the ChEMBL query:")
|
187 |
+
print(" ", str(e))
|
188 |
+
return empty_df
|
189 |
+
|
190 |
+
|
191 |
+
@op("LynxKite Graph Analytics", "get bioactivity from uniprot")
|
192 |
+
def fetch_chembl_bioactivity(*, uniprot_id: str = "Q9NZQ7"):
|
193 |
+
"""
|
194 |
+
Fetch bioactivity data from ChEMBL for a given UniProt ID.
|
195 |
+
"""
|
196 |
+
target = new_client.target.filter(target_components__accession=uniprot_id)
|
197 |
+
targets = list(target)
|
198 |
+
if not targets:
|
199 |
+
return []
|
200 |
+
|
201 |
+
target_chembl_id = targets[0]["target_chembl_id"]
|
202 |
+
activities = new_client.activity.filter(
|
203 |
+
target_chembl_id=target_chembl_id, standard_type__in=["IC50", "Ki", "Kd"]
|
204 |
+
)
|
205 |
+
df = pd.DataFrame(activities)
|
206 |
+
return df
|
examples/Cheminformatics/cheminfo_tools.py
CHANGED
@@ -16,6 +16,7 @@ from sklearn.ensemble import RandomForestRegressor
|
|
16 |
from sklearn.metrics import r2_score, mean_absolute_error, mean_squared_error
|
17 |
from sklearn.model_selection import train_test_split
|
18 |
import numpy as np
|
|
|
19 |
|
20 |
|
21 |
@op("LynxKite Graph Analytics", "View mol filter", view="matplotlib", slow=True)
|
@@ -303,3 +304,612 @@ def build_qsar_model(
|
|
303 |
|
304 |
print(f"Trained & saved QSAR model for '{fp_type}' → {model_file}")
|
305 |
return metrics_df
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
from sklearn.metrics import r2_score, mean_absolute_error, mean_squared_error
|
17 |
from sklearn.model_selection import train_test_split
|
18 |
import numpy as np
|
19 |
+
from rdkit.Chem import MACCSkeys
|
20 |
|
21 |
|
22 |
@op("LynxKite Graph Analytics", "View mol filter", view="matplotlib", slow=True)
|
|
|
304 |
|
305 |
print(f"Trained & saved QSAR model for '{fp_type}' → {model_file}")
|
306 |
return metrics_df
|
307 |
+
|
308 |
+
|
309 |
+
def predict_with_ci(model, X, confidence=0.95):
|
310 |
+
"""
|
311 |
+
Calculates predictions and confidence intervals for a RandomForestRegressor.
|
312 |
+
(Implementation is the same as in the previous answer)
|
313 |
+
"""
|
314 |
+
# Get predictions from each individual tree
|
315 |
+
tree_preds = np.array([tree.predict(X) for tree in model.estimators_])
|
316 |
+
# Calculate mean prediction
|
317 |
+
y_pred_mean = np.mean(tree_preds, axis=0)
|
318 |
+
# Calculate percentiles for confidence interval
|
319 |
+
alpha = (1.0 - confidence) / 2.0
|
320 |
+
lower_percentile = alpha * 100
|
321 |
+
upper_percentile = (1.0 - alpha) * 100
|
322 |
+
y_pred_lower = np.percentile(tree_preds, lower_percentile, axis=0)
|
323 |
+
y_pred_upper = np.percentile(tree_preds, upper_percentile, axis=0)
|
324 |
+
return y_pred_mean, y_pred_lower, y_pred_upper
|
325 |
+
|
326 |
+
|
327 |
+
# --- End of predict_with_ci definition ---
|
328 |
+
|
329 |
+
|
330 |
+
@op("LynxKite Graph Analytics", "Train QSAR2")
|
331 |
+
def build_qsar_model2(
|
332 |
+
df: pd.DataFrame,
|
333 |
+
*,
|
334 |
+
smiles_col: str,
|
335 |
+
target_col: str,
|
336 |
+
fp_type: str,
|
337 |
+
radius: int = 2,
|
338 |
+
n_bits: int = 2048,
|
339 |
+
test_size: float = 0.2,
|
340 |
+
random_state: int = 42,
|
341 |
+
out_dir: str = "Models",
|
342 |
+
confidence: float = 0.95,
|
343 |
+
):
|
344 |
+
"""
|
345 |
+
Train/save RandomForest QSAR model, returning the model and a results DataFrame.
|
346 |
+
|
347 |
+
The results DataFrame contains per-point data ('actual', 'predicted',
|
348 |
+
'lower_ci', 'upper_ci', 'split') AND repeated summary metrics for each
|
349 |
+
split ('split_R2', 'split_MAE', 'split_RMSE').
|
350 |
+
|
351 |
+
Parameters
|
352 |
+
----------
|
353 |
+
(Parameters are the same as before)
|
354 |
+
bundle : any
|
355 |
+
table_name : str
|
356 |
+
smiles_col : str
|
357 |
+
target_col : str
|
358 |
+
fp_type : str
|
359 |
+
radius : int
|
360 |
+
n_bits : int
|
361 |
+
test_size : float
|
362 |
+
random_state : int
|
363 |
+
out_dir : str
|
364 |
+
confidence : float, optional
|
365 |
+
|
366 |
+
Returns
|
367 |
+
-------
|
368 |
+
model : RandomForestRegressor
|
369 |
+
The trained QSAR model.
|
370 |
+
results_df : pandas.DataFrame
|
371 |
+
DataFrame containing columns: 'actual', 'predicted', 'lower_ci',
|
372 |
+
'upper_ci', 'split', 'split_R2', 'split_MAE', 'split_RMSE'.
|
373 |
+
The metric columns repeat the overall metric for the corresponding split.
|
374 |
+
"""
|
375 |
+
# Steps 1-5: Load data, split, featurize, split features, train model
|
376 |
+
# (Code is identical to previous versions up to model training)
|
377 |
+
# ... (load data, sanitize, split indices) ...
|
378 |
+
# df = bundle.dfs.get(table_name)
|
379 |
+
df = df.copy()
|
380 |
+
if df is None:
|
381 |
+
raise KeyError("Table not found")
|
382 |
+
df[target_col] = pd.to_numeric(df[target_col], errors="coerce")
|
383 |
+
df.dropna(subset=[target_col, smiles_col], inplace=True)
|
384 |
+
df["mol"] = df[smiles_col].apply(Chem.MolFromSmiles)
|
385 |
+
df = df[df["mol"].notnull()].reset_index(drop=True)
|
386 |
+
if df.empty:
|
387 |
+
raise ValueError("No valid molecules or targets")
|
388 |
+
|
389 |
+
indices = np.arange(len(df))
|
390 |
+
train_idx, test_idx = train_test_split(indices, test_size=test_size, random_state=random_state)
|
391 |
+
|
392 |
+
print(f"Featurizing using {fp_type}...")
|
393 |
+
fps = []
|
394 |
+
valid_indices = []
|
395 |
+
for i, mol in enumerate(df["mol"]):
|
396 |
+
try:
|
397 |
+
# ... (fp generation logic as before) ...
|
398 |
+
if fp_type == "ecfp":
|
399 |
+
bv = AllChem.GetMorganFingerprintAsBitVect(mol, radius, nBits=n_bits)
|
400 |
+
current_n_bits = n_bits
|
401 |
+
elif fp_type == "rdkit":
|
402 |
+
bv = Chem.RDKFingerprint(mol, fpSize=n_bits)
|
403 |
+
current_n_bits = n_bits
|
404 |
+
elif fp_type == "torsion":
|
405 |
+
bv = AllChem.GetHashedTopologicalTorsionFingerprintAsBitVect(mol, nBits=n_bits)
|
406 |
+
current_n_bits = n_bits
|
407 |
+
elif fp_type == "atompair":
|
408 |
+
bv = AllChem.GetHashedAtomPairFingerprintAsBitVect(mol, nBits=n_bits)
|
409 |
+
current_n_bits = n_bits
|
410 |
+
elif fp_type == "maccs":
|
411 |
+
bv = MACCSkeys.GenMACCSKeys(mol) # 167 bits
|
412 |
+
current_n_bits = 167
|
413 |
+
else:
|
414 |
+
raise ValueError(f"Unsupported fp type: '{fp_type}'")
|
415 |
+
|
416 |
+
arr = np.zeros((current_n_bits,), dtype=np.int8)
|
417 |
+
DataStructs.ConvertToNumpyArray(bv, arr)
|
418 |
+
fps.append(arr)
|
419 |
+
valid_indices.append(i)
|
420 |
+
except Exception as e:
|
421 |
+
print(f"Warning: Featurization failed index {i}. Skipping. Error: {e}")
|
422 |
+
continue
|
423 |
+
if not fps:
|
424 |
+
raise ValueError("No molecules featurized.")
|
425 |
+
X = np.vstack(fps)
|
426 |
+
df_filtered = df.iloc[valid_indices].reset_index(drop=True)
|
427 |
+
y = df_filtered[target_col].values
|
428 |
+
|
429 |
+
# original_indices_set = set(valid_indices)
|
430 |
+
|
431 |
+
train_idx_filtered = [
|
432 |
+
i for i, original_idx in enumerate(valid_indices) if original_idx in train_idx
|
433 |
+
]
|
434 |
+
test_idx_filtered = [
|
435 |
+
i for i, original_idx in enumerate(valid_indices) if original_idx in test_idx
|
436 |
+
]
|
437 |
+
|
438 |
+
X_train, y_train = X[train_idx_filtered], y[train_idx_filtered]
|
439 |
+
X_test, y_test = X[test_idx_filtered], y[test_idx_filtered]
|
440 |
+
|
441 |
+
if X_train.shape[0] == 0 or X_test.shape[0] == 0:
|
442 |
+
raise ValueError("Train or test split empty after filtering.")
|
443 |
+
|
444 |
+
print("Training RandomForestRegressor...")
|
445 |
+
model = RandomForestRegressor(random_state=random_state, n_jobs=-1)
|
446 |
+
model.fit(X_train, y_train)
|
447 |
+
|
448 |
+
# 6) Compute predictions and *summary* performance metrics
|
449 |
+
print("Calculating predictions and metrics...")
|
450 |
+
y_pred_train, lower_ci_train, upper_ci_train = predict_with_ci(model, X_train, confidence)
|
451 |
+
y_pred_test, lower_ci_test, upper_ci_test = predict_with_ci(model, X_test, confidence)
|
452 |
+
|
453 |
+
def _metrics(y_true, y_pred_mean):
|
454 |
+
# (Same helper function as before)
|
455 |
+
y_true = np.ravel(y_true)
|
456 |
+
y_pred_mean = np.ravel(y_pred_mean)
|
457 |
+
if len(y_true) == 0:
|
458 |
+
return {"R2": np.nan, "MAE": np.nan, "RMSE": np.nan}
|
459 |
+
mse = mean_squared_error(y_true, y_pred_mean)
|
460 |
+
return {
|
461 |
+
"R2": r2_score(y_true, y_pred_mean),
|
462 |
+
"MAE": mean_absolute_error(y_true, y_pred_mean),
|
463 |
+
"RMSE": np.sqrt(mse),
|
464 |
+
}
|
465 |
+
|
466 |
+
train_metrics_dict = _metrics(y_train, y_pred_train)
|
467 |
+
test_metrics_dict = _metrics(y_test, y_pred_test)
|
468 |
+
|
469 |
+
# 7) Create results DataFrames and ADD metrics columns
|
470 |
+
train_results = pd.DataFrame(
|
471 |
+
{
|
472 |
+
"actual": y_train,
|
473 |
+
"predicted": y_pred_train,
|
474 |
+
"lower_ci": lower_ci_train,
|
475 |
+
"upper_ci": upper_ci_train,
|
476 |
+
"split": "train",
|
477 |
+
}
|
478 |
+
)
|
479 |
+
# Add repeated metrics
|
480 |
+
for metric, value in train_metrics_dict.items():
|
481 |
+
train_results[f"split_{metric}"] = value
|
482 |
+
|
483 |
+
test_results = pd.DataFrame(
|
484 |
+
{
|
485 |
+
"actual": y_test,
|
486 |
+
"predicted": y_pred_test,
|
487 |
+
"lower_ci": lower_ci_test,
|
488 |
+
"upper_ci": upper_ci_test,
|
489 |
+
"split": "test",
|
490 |
+
}
|
491 |
+
)
|
492 |
+
# Add repeated metrics
|
493 |
+
for metric, value in test_metrics_dict.items():
|
494 |
+
test_results[f"split_{metric}"] = value
|
495 |
+
|
496 |
+
# Concatenate into the final DataFrame
|
497 |
+
results_df = pd.concat([train_results, test_results], ignore_index=True)
|
498 |
+
|
499 |
+
# 8) Save the model (same as before)
|
500 |
+
os.makedirs(out_dir, exist_ok=True)
|
501 |
+
model_file = os.path.join(out_dir, f"qsar_model_{fp_type}.pkl")
|
502 |
+
try:
|
503 |
+
with open(model_file, "wb") as fout:
|
504 |
+
pickle.dump(model, fout)
|
505 |
+
print(f"Trained & saved QSAR model for '{fp_type}' -> {model_file}")
|
506 |
+
except Exception as e:
|
507 |
+
print(f"Error saving model to {model_file}: {e}")
|
508 |
+
|
509 |
+
return results_df
|
510 |
+
|
511 |
+
|
512 |
+
@op("LynxKite Graph Analytics", "plot qsar", view="matplotlib")
|
513 |
+
def plot_qsar(results_df: pd.DataFrame):
|
514 |
+
"""
|
515 |
+
Plots actual vs. predicted values from a QSAR results DataFrame.
|
516 |
+
|
517 |
+
Requires a single positional argument: the results DataFrame. All other
|
518 |
+
parameters are optional keyword arguments. It extracts summary metrics
|
519 |
+
directly from columns ('split_R2', 'split_MAE', 'split_RMSE')
|
520 |
+
expected within the results_df.
|
521 |
+
"""
|
522 |
+
title = "QSAR Model Performance: Actual vs. Predicted"
|
523 |
+
xlabel = "Actual Values"
|
524 |
+
ylabel = "Predicted Values"
|
525 |
+
show_metrics = True
|
526 |
+
|
527 |
+
if not isinstance(results_df, pd.DataFrame):
|
528 |
+
raise TypeError(
|
529 |
+
"plot_qsar() missing 1 required positional argument: 'results_df' or the provided argument is not a pandas DataFrame."
|
530 |
+
)
|
531 |
+
|
532 |
+
required_cols = ["actual", "predicted", "lower_ci", "upper_ci", "split"]
|
533 |
+
if not all(col in results_df.columns for col in required_cols):
|
534 |
+
raise ValueError(f"Invalid 'results_df'. Must contain columns: {required_cols}")
|
535 |
+
|
536 |
+
metric_cols = ["split_R2", "split_MAE", "split_RMSE"]
|
537 |
+
metrics_available = all(col in results_df.columns for col in metric_cols)
|
538 |
+
if show_metrics and not metrics_available:
|
539 |
+
print(
|
540 |
+
f"Warning: Metrics display requested, but one or more metric columns ({metric_cols}) are missing in results_df."
|
541 |
+
)
|
542 |
+
|
543 |
+
# --- Prepare Data ---
|
544 |
+
train_data = results_df[results_df["split"] == "train"]
|
545 |
+
test_data = results_df[results_df["split"] == "test"]
|
546 |
+
can_plot_train = not train_data.empty
|
547 |
+
can_plot_test = not test_data.empty
|
548 |
+
|
549 |
+
if not can_plot_train and not can_plot_test:
|
550 |
+
print("Warning: Both training and test data subsets are empty. Cannot generate plot.")
|
551 |
+
return # Exit function early if no data
|
552 |
+
|
553 |
+
# --- Create Plot (Internal Figure/Axes) ---
|
554 |
+
fig, ax = plt.subplots(figsize=(8, 8))
|
555 |
+
|
556 |
+
# --- Plotting Logic ---
|
557 |
+
# (Draws scatter, error bars, line, grid, labels, title, legend on 'ax')
|
558 |
+
if can_plot_train:
|
559 |
+
train_error = [
|
560 |
+
train_data["predicted"] - train_data["lower_ci"],
|
561 |
+
train_data["upper_ci"] - train_data["predicted"],
|
562 |
+
]
|
563 |
+
ax.scatter(
|
564 |
+
train_data["actual"],
|
565 |
+
train_data["predicted"],
|
566 |
+
label="Train",
|
567 |
+
alpha=0.6,
|
568 |
+
s=30,
|
569 |
+
edgecolors="w",
|
570 |
+
linewidth=0.5,
|
571 |
+
)
|
572 |
+
ax.errorbar(
|
573 |
+
train_data["actual"],
|
574 |
+
train_data["predicted"],
|
575 |
+
yerr=train_error,
|
576 |
+
fmt="none",
|
577 |
+
ecolor="tab:blue",
|
578 |
+
label="_nolegend_",
|
579 |
+
capsize=0,
|
580 |
+
elinewidth=1,
|
581 |
+
)
|
582 |
+
|
583 |
+
if can_plot_test:
|
584 |
+
test_error = [
|
585 |
+
test_data["predicted"] - test_data["lower_ci"],
|
586 |
+
test_data["upper_ci"] - test_data["predicted"],
|
587 |
+
]
|
588 |
+
ax.scatter(
|
589 |
+
test_data["actual"],
|
590 |
+
test_data["predicted"],
|
591 |
+
label="Test",
|
592 |
+
alpha=0.8,
|
593 |
+
s=40,
|
594 |
+
edgecolors="w",
|
595 |
+
linewidth=0.5,
|
596 |
+
)
|
597 |
+
ax.errorbar(
|
598 |
+
test_data["actual"],
|
599 |
+
test_data["predicted"],
|
600 |
+
yerr=test_error,
|
601 |
+
fmt="none",
|
602 |
+
ecolor="tab:orange",
|
603 |
+
label="_nolegend_",
|
604 |
+
capsize=0,
|
605 |
+
elinewidth=1,
|
606 |
+
)
|
607 |
+
|
608 |
+
all_actual = results_df["actual"].dropna()
|
609 |
+
all_pred_ci = pd.concat(
|
610 |
+
[results_df["predicted"], results_df["lower_ci"], results_df["upper_ci"]]
|
611 |
+
).dropna()
|
612 |
+
all_values = pd.concat([all_actual, all_pred_ci]).dropna()
|
613 |
+
if all_values.empty:
|
614 |
+
min_val, max_val = 0, 1
|
615 |
+
else:
|
616 |
+
min_val, max_val = all_values.min(), all_values.max()
|
617 |
+
if min_val == max_val:
|
618 |
+
min_val -= 0.5
|
619 |
+
max_val += 0.5
|
620 |
+
padding = (max_val - min_val) * 0.05
|
621 |
+
min_val -= padding
|
622 |
+
max_val += padding
|
623 |
+
ax.plot([min_val, max_val], [min_val, max_val], "k--", alpha=0.7, lw=1, label="y=x")
|
624 |
+
ax.set_xlim(min_val, max_val)
|
625 |
+
ax.set_ylim(min_val, max_val)
|
626 |
+
ax.set_aspect("equal", adjustable="box")
|
627 |
+
ax.grid(True, linestyle=":", alpha=0.6)
|
628 |
+
ax.set_xlabel(xlabel)
|
629 |
+
ax.set_ylabel(ylabel)
|
630 |
+
ax.set_title(title)
|
631 |
+
ax.legend(loc="lower right")
|
632 |
+
|
633 |
+
# --- Display Metrics Text ---
|
634 |
+
if show_metrics and metrics_available:
|
635 |
+
# (Logic for extracting and formatting metrics text remains the same)
|
636 |
+
metrics_text = ""
|
637 |
+
try:
|
638 |
+
if can_plot_train:
|
639 |
+
train_metrics = train_data[metric_cols].iloc[0]
|
640 |
+
r2_tr = (
|
641 |
+
f"{train_metrics['split_R2']:.3f}"
|
642 |
+
if pd.notna(train_metrics["split_R2"])
|
643 |
+
else "N/A"
|
644 |
+
)
|
645 |
+
mae_tr = (
|
646 |
+
f"{train_metrics['split_MAE']:.3f}"
|
647 |
+
if pd.notna(train_metrics["split_MAE"])
|
648 |
+
else "N/A"
|
649 |
+
)
|
650 |
+
rmse_tr = (
|
651 |
+
f"{train_metrics['split_RMSE']:.3f}"
|
652 |
+
if pd.notna(train_metrics["split_RMSE"])
|
653 |
+
else "N/A"
|
654 |
+
)
|
655 |
+
metrics_text += f"Train: $R^2$={r2_tr}, MAE={mae_tr}, RMSE={rmse_tr}\n"
|
656 |
+
else:
|
657 |
+
metrics_text += "Train: N/A (No Data)\n"
|
658 |
+
if can_plot_test:
|
659 |
+
test_metrics = test_data[metric_cols].iloc[0]
|
660 |
+
r2_te = (
|
661 |
+
f"{test_metrics['split_R2']:.3f}"
|
662 |
+
if pd.notna(test_metrics["split_R2"])
|
663 |
+
else "N/A"
|
664 |
+
)
|
665 |
+
mae_te = (
|
666 |
+
f"{test_metrics['split_MAE']:.3f}"
|
667 |
+
if pd.notna(test_metrics["split_MAE"])
|
668 |
+
else "N/A"
|
669 |
+
)
|
670 |
+
rmse_te = (
|
671 |
+
f"{test_metrics['split_RMSE']:.3f}"
|
672 |
+
if pd.notna(test_metrics["split_RMSE"])
|
673 |
+
else "N/A"
|
674 |
+
)
|
675 |
+
metrics_text += f"Test: $R^2$={r2_te}, MAE={mae_te}, RMSE={rmse_te}"
|
676 |
+
else:
|
677 |
+
metrics_text += "Test: N/A (No Data)"
|
678 |
+
if metrics_text:
|
679 |
+
ax.text(
|
680 |
+
0.05,
|
681 |
+
0.95,
|
682 |
+
metrics_text.strip(),
|
683 |
+
transform=ax.transAxes,
|
684 |
+
fontsize=9,
|
685 |
+
verticalalignment="top",
|
686 |
+
bbox=dict(boxstyle="round,pad=0.5", fc="white", alpha=0.8),
|
687 |
+
)
|
688 |
+
except Exception as e:
|
689 |
+
print(f"An error occurred during metrics display: {e}")
|
690 |
+
ax.text(
|
691 |
+
0.05,
|
692 |
+
0.95,
|
693 |
+
"Error displaying metrics",
|
694 |
+
transform=ax.transAxes,
|
695 |
+
fontsize=9,
|
696 |
+
color="red",
|
697 |
+
verticalalignment="top",
|
698 |
+
bbox=dict(boxstyle="round,pad=0.5", fc="white", alpha=0.8),
|
699 |
+
)
|
700 |
+
|
701 |
+
|
702 |
+
@op("LynxKite Graph Analytics", "plot qsar2", view="matplotlib")
|
703 |
+
def plot_qsar2(results_df: pd.DataFrame):
|
704 |
+
"""
|
705 |
+
Plots actual vs. predicted values resembling the example image.
|
706 |
+
|
707 |
+
Includes separate markers for train/test, y=x line, and parallel dashed
|
708 |
+
error bands based on test set RMSE (optional). Does NOT use per-point CIs.
|
709 |
+
|
710 |
+
Handles displaying the plot via plt.show() or saving it to a file
|
711 |
+
based on the `save_path` parameter. THIS FUNCTION DOES NOT RETURN ANY VALUE.
|
712 |
+
|
713 |
+
Parameters
|
714 |
+
----------
|
715 |
+
results_df : pd.DataFrame
|
716 |
+
Mandatory input DataFrame. Must contain: 'actual', 'predicted', 'split'.
|
717 |
+
Should also contain 'split_RMSE' column for error bands and metrics display.
|
718 |
+
title : str, optional
|
719 |
+
xlabel : str, optional
|
720 |
+
ylabel : str, optional
|
721 |
+
rmse_multiplier_for_bands : float or None, optional
|
722 |
+
Determines the width of the dashed error bands (multiplier * test_RMSE).
|
723 |
+
Set to None to disable bands. Default is 1.0.
|
724 |
+
show_metrics : bool, optional
|
725 |
+
Whether to display R2/MAE/RMSE text (requires metric columns). Default is True.
|
726 |
+
save_path : str, optional
|
727 |
+
If provided, saves plot to this path. If None (default), displays plot.
|
728 |
+
|
729 |
+
Raises
|
730 |
+
------
|
731 |
+
ValueError / TypeError : For invalid inputs.
|
732 |
+
"""
|
733 |
+
COLOR_TRAIN = "royalblue"
|
734 |
+
COLOR_TEST = "darkorange" # Changed from red for potentially better contrast/appeal
|
735 |
+
COLOR_PERFECT = "black"
|
736 |
+
COLOR_BANDS = "dimgrey" # Less prominent than the perfect line
|
737 |
+
COLOR_GRID = "lightgrey"
|
738 |
+
title = "QSAR Model Performance: Actual vs. Predicted"
|
739 |
+
xlabel = "Actual Values"
|
740 |
+
ylabel = "Predicted Values"
|
741 |
+
# ci_alpha = 0.2
|
742 |
+
show_metrics = True
|
743 |
+
rmse_multiplier_for_bands = 1.0
|
744 |
+
# --- Input Validation ---
|
745 |
+
if not isinstance(results_df, pd.DataFrame):
|
746 |
+
raise TypeError("Input must be a pandas DataFrame.")
|
747 |
+
|
748 |
+
required_cols = ["actual", "predicted", "split"]
|
749 |
+
if not all(col in results_df.columns for col in required_cols):
|
750 |
+
raise ValueError(f"DataFrame must contain columns: {required_cols}")
|
751 |
+
|
752 |
+
metric_cols = ["split_R2", "split_MAE", "split_RMSE"]
|
753 |
+
metrics_available = all(col in results_df.columns for col in metric_cols)
|
754 |
+
bands_possible = rmse_multiplier_for_bands is not None and "split_RMSE" in results_df.columns
|
755 |
+
|
756 |
+
if show_metrics and not metrics_available:
|
757 |
+
print(
|
758 |
+
f"Warning: Metrics display requested, but one or more metric columns ({metric_cols}) are missing."
|
759 |
+
)
|
760 |
+
if rmse_multiplier_for_bands is not None and "split_RMSE" not in results_df.columns:
|
761 |
+
print("Warning: Error bands requested, but 'split_RMSE' column is missing.")
|
762 |
+
bands_possible = False
|
763 |
+
|
764 |
+
# --- Prepare Data ---
|
765 |
+
train_data = results_df[results_df["split"] == "train"].copy()
|
766 |
+
test_data = results_df[results_df["split"] == "test"].copy()
|
767 |
+
can_plot_train = not train_data.empty
|
768 |
+
can_plot_test = not test_data.empty
|
769 |
+
|
770 |
+
if not can_plot_train and not can_plot_test:
|
771 |
+
print("Warning: Both training and test data subsets are empty. Cannot generate plot.")
|
772 |
+
return
|
773 |
+
|
774 |
+
# --- Create Plot with Style ---
|
775 |
+
plt.style.use("seaborn-v0_8-whitegrid") # Use a cleaner base style
|
776 |
+
fig, ax = plt.subplots(figsize=(8, 8)) # Slightly larger figure
|
777 |
+
|
778 |
+
# --- Plotting Logic ---
|
779 |
+
# Scatter plots with enhanced style
|
780 |
+
common_scatter_kws = {"s": 45, "alpha": 0.75, "edgecolor": "black", "linewidth": 0.5}
|
781 |
+
if can_plot_train:
|
782 |
+
ax.scatter(
|
783 |
+
train_data["actual"],
|
784 |
+
train_data["predicted"],
|
785 |
+
label="Training set",
|
786 |
+
marker="o",
|
787 |
+
color=COLOR_TRAIN,
|
788 |
+
**common_scatter_kws,
|
789 |
+
) # Blue circles
|
790 |
+
|
791 |
+
if can_plot_test:
|
792 |
+
ax.scatter(
|
793 |
+
test_data["actual"],
|
794 |
+
test_data["predicted"],
|
795 |
+
label="Test set",
|
796 |
+
marker="o",
|
797 |
+
color=COLOR_TEST,
|
798 |
+
**common_scatter_kws,
|
799 |
+
) # Orange circles
|
800 |
+
|
801 |
+
# Determine plot limits
|
802 |
+
# (Using the same logic as before to calculate min_val, max_val)
|
803 |
+
all_actual = results_df["actual"].dropna()
|
804 |
+
all_pred = results_df["predicted"].dropna()
|
805 |
+
all_values = pd.concat([all_actual, all_pred]).dropna()
|
806 |
+
if all_values.empty:
|
807 |
+
min_val, max_val = 0, 1
|
808 |
+
else:
|
809 |
+
min_val, max_val = all_values.min(), all_values.max()
|
810 |
+
if min_val == max_val:
|
811 |
+
min_val -= 0.5
|
812 |
+
max_val += 0.5
|
813 |
+
data_range = max_val - min_val
|
814 |
+
if data_range == 0:
|
815 |
+
data_range = 1.0
|
816 |
+
padding = data_range * 0.10
|
817 |
+
min_val -= padding
|
818 |
+
max_val += padding
|
819 |
+
|
820 |
+
# Plot y=x line (Solid Black, slightly thicker)
|
821 |
+
ax.plot(
|
822 |
+
[min_val, max_val],
|
823 |
+
[min_val, max_val],
|
824 |
+
color=COLOR_PERFECT,
|
825 |
+
linestyle="-",
|
826 |
+
linewidth=1.5,
|
827 |
+
alpha=0.9,
|
828 |
+
label="_nolegend_",
|
829 |
+
)
|
830 |
+
|
831 |
+
# Plot Error Bands based on Test RMSE (subtler style)
|
832 |
+
rmse_test = np.nan
|
833 |
+
if bands_possible and can_plot_test:
|
834 |
+
try:
|
835 |
+
rmse_test = test_data["split_RMSE"].dropna().iloc[0]
|
836 |
+
if pd.notna(rmse_test) and rmse_test >= 0:
|
837 |
+
margin = rmse_multiplier_for_bands * rmse_test
|
838 |
+
band_label = (
|
839 |
+
f"$\pm {rmse_multiplier_for_bands}\,$RMSE"
|
840 |
+
if rmse_multiplier_for_bands == 1
|
841 |
+
else f"$\pm {rmse_multiplier_for_bands}\,$RMSE"
|
842 |
+
)
|
843 |
+
ax.plot(
|
844 |
+
[min_val, max_val],
|
845 |
+
[min_val + margin, max_val + margin],
|
846 |
+
color=COLOR_BANDS,
|
847 |
+
linestyle="--",
|
848 |
+
linewidth=1.0,
|
849 |
+
alpha=0.7,
|
850 |
+
label=band_label,
|
851 |
+
) # Grey dashed
|
852 |
+
ax.plot(
|
853 |
+
[min_val, max_val],
|
854 |
+
[min_val - margin, max_val - margin],
|
855 |
+
color=COLOR_BANDS,
|
856 |
+
linestyle="--",
|
857 |
+
linewidth=1.0,
|
858 |
+
alpha=0.7,
|
859 |
+
label="_nolegend_",
|
860 |
+
) # Grey dashed
|
861 |
+
# else: print("Warning: Could not plot error bands (Invalid Test RMSE).") # Optionally silent
|
862 |
+
except Exception as e:
|
863 |
+
print(f"Warning: Could not plot error bands: {e}")
|
864 |
+
|
865 |
+
# Set limits and aspect ratio
|
866 |
+
ax.set_xlim(min_val, max_val)
|
867 |
+
ax.set_ylim(min_val, max_val)
|
868 |
+
ax.set_aspect("equal", adjustable="box")
|
869 |
+
|
870 |
+
# ADD BACK Grid (Subtle Style)
|
871 |
+
ax.grid(True, which="both", linestyle=":", linewidth=0.7, color=COLOR_GRID, alpha=0.7)
|
872 |
+
# Ensure grid is behind data points
|
873 |
+
ax.set_axisbelow(True)
|
874 |
+
|
875 |
+
# Set Labels and Title (using specified arguments)
|
876 |
+
ax.set_xlabel(xlabel, fontsize=12)
|
877 |
+
ax.set_ylabel(ylabel, fontsize=12)
|
878 |
+
ax.set_title(title, fontsize=15, pad=15, weight="semibold") # Slightly larger title
|
879 |
+
|
880 |
+
# Enhance Legend
|
881 |
+
ax.legend(loc="best", frameon=True, framealpha=0.85, fontsize=10, shadow=False)
|
882 |
+
|
883 |
+
# --- Display Metrics Text (Optional) ---
|
884 |
+
if show_metrics and metrics_available:
|
885 |
+
# (Logic for extracting and formatting metrics text remains the same)
|
886 |
+
metrics_text = ""
|
887 |
+
try:
|
888 |
+
if can_plot_train:
|
889 |
+
train_metrics = train_data[metric_cols].dropna().iloc[0] # Ensure using valid row
|
890 |
+
r2_tr = f"{train_metrics['split_R2']:.3f}"
|
891 |
+
mae_tr = f"{train_metrics['split_MAE']:.3f}"
|
892 |
+
rmse_tr = f"{train_metrics['split_RMSE']:.3f}"
|
893 |
+
metrics_text += f"Train: $R^2$={r2_tr}, MAE={mae_tr}, RMSE={rmse_tr}\n"
|
894 |
+
else:
|
895 |
+
metrics_text += "Train: N/A\n"
|
896 |
+
if can_plot_test:
|
897 |
+
test_metrics = test_data[metric_cols].dropna().iloc[0] # Ensure using valid row
|
898 |
+
r2_te = f"{test_metrics['split_R2']:.3f}"
|
899 |
+
mae_te = f"{test_metrics['split_MAE']:.3f}"
|
900 |
+
rmse_te = f"{test_metrics['split_RMSE']:.3f}"
|
901 |
+
metrics_text += f"Test: $R^2$={r2_te}, MAE={mae_te}, RMSE={rmse_te}"
|
902 |
+
else:
|
903 |
+
metrics_text += "Test: N/A"
|
904 |
+
if metrics_text:
|
905 |
+
ax.text(
|
906 |
+
0.05,
|
907 |
+
0.95,
|
908 |
+
metrics_text.strip(),
|
909 |
+
transform=ax.transAxes,
|
910 |
+
fontsize=9,
|
911 |
+
verticalalignment="top",
|
912 |
+
bbox=dict(boxstyle="round,pad=0.3", fc="white", alpha=0.7),
|
913 |
+
) # Adjusted box slightly
|
914 |
+
except Exception as e:
|
915 |
+
print(f"An error occurred during metrics display: {e}")
|
examples/Cheminformatics/qsar_example.lynxkite.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
examples/draw_molecules.py
DELETED
@@ -1,29 +0,0 @@
|
|
1 |
-
from lynxkite.core.ops import op
|
2 |
-
import pandas as pd
|
3 |
-
import base64
|
4 |
-
import io
|
5 |
-
|
6 |
-
|
7 |
-
def pil_to_data(image):
|
8 |
-
buffer = io.BytesIO()
|
9 |
-
image.save(buffer, format="png")
|
10 |
-
b64 = base64.b64encode(buffer.getvalue()).decode("utf-8")
|
11 |
-
return "data:image/png;base64," + b64
|
12 |
-
|
13 |
-
|
14 |
-
def smiles_to_data(smiles):
|
15 |
-
import rdkit
|
16 |
-
|
17 |
-
m = rdkit.Chem.MolFromSmiles(smiles)
|
18 |
-
if m is None:
|
19 |
-
return None
|
20 |
-
img = rdkit.Chem.Draw.MolToImage(m)
|
21 |
-
data = pil_to_data(img)
|
22 |
-
return data
|
23 |
-
|
24 |
-
|
25 |
-
@op("LynxKite Graph Analytics", "Draw molecules")
|
26 |
-
def draw_molecules(df: pd.DataFrame, *, smiles_column: str, image_column: str = "image"):
|
27 |
-
df = df.copy()
|
28 |
-
df[image_column] = df[smiles_column].apply(smiles_to_data)
|
29 |
-
return df
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
examples/requirements.txt
CHANGED
@@ -1,3 +1,6 @@
|
|
1 |
# Example of a requirements.txt file. LynxKite will automatically install anything you put here.
|
2 |
faker
|
3 |
matplotlib
|
|
|
|
|
|
|
|
1 |
# Example of a requirements.txt file. LynxKite will automatically install anything you put here.
|
2 |
faker
|
3 |
matplotlib
|
4 |
+
chembl_webresource_client
|
5 |
+
rcsb-api
|
6 |
+
itertools
|