File size: 13,244 Bytes
0eb933d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 |
#!/usr/bin/env python3
"""
method_split.py
Author: natelgrw
Last Edited: 11/04/2025
Splits the ReTiNA dataset into 5 folds based on LC-MS setup configurations
(e.g., solvents, gradient, column, temperature, flow rate).
Each setup group is assigned to a single fold to avoid data leakage
across similar chromatographic conditions.
"""
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from pathlib import Path
from collections import defaultdict
import random
import umap
from sklearn.preprocessing import StandardScaler, OneHotEncoder
from sklearn.compose import ColumnTransformer
from sklearn.cluster import KMeans
from ast import literal_eval
# ===== Configuration ===== #
INPUT_CSV = "../retina_dataset.csv"
METHOD_CSV = "../lcms_methods.csv"
OUTPUT_DIR = "../method_split"
N_FOLDS = 5
RANDOM_SEED = 42
random.seed(RANDOM_SEED)
np.random.seed(RANDOM_SEED)
# ===== Helper Functions ===== #
def analyze_dataset(df):
"""
Prints dataset statistics.
"""
print("=" * 70)
print("Dataset Analysis")
print("=" * 70)
print(f"Total rows: {len(df):,}")
if "compound" in df.columns:
print(f"Unique compounds: {df['compound'].nunique():,}")
if 'rt' in df.columns:
print(f"\nRetention Time Stats:")
print(f" Mean: {df['rt'].mean():.2f} s | Median: {df['rt'].median():.2f} s")
if 'method_number' in df.columns:
print(f"Unique methods: {df['method_number'].nunique():,}")
print()
def assign_methods_to_folds(method_sizes, n_folds, method_features=None, method_ids=None):
"""
Assigns methods to folds using KMeans clustering on UMAP coordinates.
This creates compact, spatially-separated regions in the UMAP visualization.
"""
fold_assignments = defaultdict(list)
fold_counts = [0] * n_folds
if method_features is not None and method_ids is not None:
print(f"\nUsing KMeans to create {n_folds} spatially-separated regions...")
kmeans = KMeans(n_clusters=n_folds, random_state=RANDOM_SEED, n_init=20)
cluster_labels = kmeans.fit_predict(method_features)
cluster_to_methods = defaultdict(list)
for method_id, cluster_id in zip(method_ids, cluster_labels):
cluster_to_methods[cluster_id].append(method_id)
cluster_sizes = {}
for cluster_id, methods in cluster_to_methods.items():
cluster_sizes[cluster_id] = sum(method_sizes[m] for m in methods)
total_size = sum(method_sizes.values())
avg_size = total_size / n_folds
print(f"Target size per fold: {avg_size:,.0f} datapoints")
for cluster_id, methods in cluster_to_methods.items():
cluster_size = cluster_sizes[cluster_id]
method_size_list = [(m, method_sizes[m]) for m in methods]
method_size_list.sort(key=lambda x: x[1], reverse=True)
if len(method_size_list) > 0:
largest_method, largest_size = method_size_list[0]
if largest_size > cluster_size * 0.6 and largest_size > avg_size * 0.5:
print(f"Warning: Method {largest_method} has {largest_size:,} datapoints "
f"({100*largest_size/total_size:.1f}% of total dataset)")
for cluster_id in range(n_folds):
methods = cluster_to_methods[cluster_id]
for method in methods:
fold_assignments[cluster_id].append(method)
fold_counts[cluster_id] += method_sizes[method]
print("\nSpatial assignment complete!")
print(f"Largest fold: {max(fold_counts):,} datapoints")
print(f"Smallest fold: {min(fold_counts):,} datapoints")
print(f"Fold size ratio: {max(fold_counts)/min(fold_counts):.2f}x")
else:
print("\nWarning: No method features provided, using greedy assignment")
sorted_methods = sorted(method_sizes.items(), key=lambda x: x[1], reverse=True)
for method, size in sorted_methods:
min_fold = min(range(n_folds), key=lambda i: fold_counts[i])
fold_assignments[min_fold].append(method)
fold_counts[min_fold] += size
return fold_assignments, fold_counts
# ===== Analyzer Class ===== #
class MethodAnalyzer:
"""
Analyzer for LC-MS setup-based splitting.
"""
def __init__(self, data_path, method_path, output_dir):
self.data_path = Path(data_path)
self.method_path = Path(method_path)
self.output_dir = Path(output_dir)
self.output_dir.mkdir(exist_ok=True, parents=True)
self.df = None
self.methods = None
def load_data(self):
"""
Loads the datasets.
"""
print("\nLOADING RETINA + METHOD DATA")
self.df = pd.read_csv(self.data_path)
self.methods = pd.read_csv(self.method_path)
for col in ["gradient", "column"]:
if col in self.methods.columns:
self.methods[col] = self.methods[col].apply(lambda x: literal_eval(x) if isinstance(x, str) else x)
print(f"Loaded {len(self.df):,} datapoints and {len(self.methods):,} methods.")
def extract_method_features(self):
"""
Converts LC-MS setups into numerical feature vectors.
"""
if hasattr(self, 'method_features_X'):
return self.method_features_X
df = self.methods.copy()
df["phase"] = df["column"].apply(lambda x: x[0] if isinstance(x, list) else "UNK")
df["col_diam"] = df["column"].apply(lambda x: float(x[1]) if isinstance(x, list) else 0)
df["col_len"] = df["column"].apply(lambda x: float(x[2]) if isinstance(x, list) else 0)
df["col_part"] = df["column"].apply(lambda x: float(x[3]) if isinstance(x, list) else 0)
df["grad_len"] = df["gradient"].apply(lambda x: x[-1][0] if isinstance(x, list) and len(x) > 0 else 0)
df["grad_range"] = df["gradient"].apply(
lambda x: max([p[1] for p in x]) - min([p[1] for p in x]) if isinstance(x, list) and len(x) > 0 else 0
)
feat_cols = ["phase", "col_diam", "col_len", "col_part", "grad_len", "grad_range", "flow_rate", "temp"]
cat_cols = ["phase"]
num_cols = [c for c in feat_cols if c not in cat_cols]
preprocessor = ColumnTransformer([
("cat", OneHotEncoder(), cat_cols),
("num", StandardScaler(), num_cols)
])
X = preprocessor.fit_transform(df[feat_cols])
df["vector"] = list(X.toarray() if hasattr(X, "toarray") else X)
print(f"Extracted {X.shape[1]} features per method.")
self.methods = df
self.method_features_X = X
return X
def create_method_splits(self, n_splits=5):
"""
Splits the dataset by LC-MS setup groups into folds.
"""
print("\nCREATING METHOD SPLITS")
print("=" * 60)
print("Extracting method features and computing UMAP...")
X = self.extract_method_features()
# compute UMAP embedding
reducer = umap.UMAP(
n_neighbors=8,
min_dist=0.1,
metric="euclidean",
random_state=RANDOM_SEED
)
umap_embedding = reducer.fit_transform(X)
self.methods["umap_x"], self.methods["umap_y"] = umap_embedding[:, 0], umap_embedding[:, 1]
print("UMAP embedding computed.")
method_ids = self.methods["method_number"].values
method_sizes = self.df["method_number"].value_counts().to_dict()
print(f"{len(method_sizes):,} unique methods in dataset.")
fold_assignments, fold_counts = assign_methods_to_folds(
method_sizes, n_splits,
method_features=umap_embedding,
method_ids=method_ids
)
print("\nFold balance summary:")
for i, count in enumerate(fold_counts):
print(f" Fold {i+1}: {count:,} datapoints ({100*count/len(self.df):.2f}%)")
method_to_fold = {}
for i, methods in fold_assignments.items():
for m in methods:
method_to_fold[m] = i + 1
self.df["fold"] = self.df["method_number"].map(method_to_fold)
self.methods["fold"] = self.methods["method_number"].map(method_to_fold)
fold_dataframes = {}
for i in range(n_splits):
fold_df = self.df[self.df["fold"] == i + 1].copy()
out_file = self.output_dir / f"methods_{i+1}.csv"
fold_df.to_csv(out_file, index=False)
fold_dataframes[i] = fold_df
print(f"Saved setup_{i+1}.csv ({len(fold_df):,} rows)")
return fold_dataframes
def visualize_rt_distributions(self, fold_dataframes):
"""
Generates a KDE plot of the RT distribution per setup split.
"""
print("\nPLOTTING RETENTION TIME DISTRIBUTIONS")
fig, ax = plt.subplots(figsize=(14, 6))
colors = sns.color_palette("husl", len(fold_dataframes))
if "rt" in self.df.columns:
overall_rt = self.df["rt"].dropna() / 60.0
if len(overall_rt) > 0:
sns.kdeplot(
overall_rt, ax=ax,
color='black', linewidth=2.5,
linestyle='--',
label=f"Overall (n={len(overall_rt):,})"
)
for i, fold_df in fold_dataframes.items():
if "rt" not in fold_df.columns:
continue
rt_min = fold_df["rt"].dropna() / 60.0
if len(rt_min) > 0:
sns.kdeplot(
rt_min, ax=ax,
label=f"Setup {i+1} (n={len(rt_min):,})",
color=colors[i],
linewidth=2.5
)
ax.set_xlabel("Retention Time (min)", fontsize=12, fontweight='bold')
ax.set_ylabel("Density", fontsize=12, fontweight='bold')
ax.set_title("Retention Time Distribution Across Method Splits", fontsize=14, fontweight='bold')
ax.legend(fontsize=10, framealpha=0.9)
ax.grid(alpha=0.3, linestyle=':', linewidth=0.5)
ax.set_xlim(left=0)
fig_dir = self.output_dir / "figures"
fig_dir.mkdir(exist_ok=True)
plt.savefig(fig_dir / "methods_rt.png", dpi=300, bbox_inches="tight")
plt.close()
print(f"Saved RT KDE plot to figures/method_rt.png")
def generate_umap_plot(self):
"""
Generates a UMAP visualization plot (embedding already computed).
"""
print("\nGENERATING UMAP VISUALIZATION")
if "umap_x" not in self.methods.columns or "umap_y" not in self.methods.columns:
print("Warning: UMAP coordinates not found, computing now...")
X = self.extract_method_features()
reducer = umap.UMAP(
n_neighbors=8,
min_dist=0.1,
metric="euclidean",
random_state=RANDOM_SEED
)
embedding = reducer.fit_transform(X)
self.methods["umap_x"], self.methods["umap_y"] = embedding[:, 0], embedding[:, 1]
plt.figure(figsize=(10, 7))
if "fold" in self.methods.columns:
colors = sns.color_palette("husl", N_FOLDS)
for fold_num in sorted(self.methods["fold"].dropna().unique()):
fold_data = self.methods[self.methods["fold"] == fold_num]
n_methods = len(fold_data)
plt.scatter(
fold_data["umap_x"], fold_data["umap_y"],
label=f"Cluster {int(fold_num)} (n={n_methods} methods)",
s=100, alpha=0.7, edgecolor="k", linewidth=0.5,
color=colors[int(fold_num) - 1]
)
plt.legend(bbox_to_anchor=(1.05, 1), loc='upper left', fontsize=10)
plt.title("UMAP Projection of LC-MS Method Space (Colored by Method Split)", fontsize=14, pad=15, fontweight='bold')
else:
sns.scatterplot(
x="umap_x", y="umap_y", data=self.methods,
s=70, color="steelblue", edgecolor="k"
)
plt.title("2D UMAP Visualization of LC-MS Method Space", fontsize=14, pad=15)
plt.tight_layout()
fig_dir = self.output_dir / "figures"
fig_dir.mkdir(exist_ok=True)
plt.savefig(fig_dir / "methods_umap.png", dpi=300, bbox_inches="tight")
plt.close()
print(f"Saved UMAP plot to figures/method_umap.png")
# ===== Main ===== #
def main():
"""
Main execution function.
"""
print("\n" + "=" * 80)
print(" " * 30 + "METHOD SPLIT PIPELINE")
print("=" * 80)
analyzer = MethodAnalyzer(INPUT_CSV, METHOD_CSV, OUTPUT_DIR)
analyzer.load_data()
analyze_dataset(analyzer.df)
fold_dataframes = analyzer.create_method_splits(n_splits=N_FOLDS)
analyzer.generate_umap_plot()
analyzer.visualize_rt_distributions(fold_dataframes)
print("\n" + "=" * 80)
print(" " * 30 + "METHOD SPLIT COMPLETE!")
print("=" * 80)
if __name__ == "__main__":
main() |