Tutorials

Interactive notebooks demonstrating MaldiDeepKit workflows end to end. Each notebook is self-contained and can be downloaded from the GitHub repository.

• 01 - Quick start
  • 1. Load the demo dataset
  • 2. Fit a classifier
  • 3. Save and reload
• 02 - Model comparison
  • Mean-of-probabilities ensemble
• 03 - Attention interpretation
• 04 - Full pipeline
  • Cross-validation with on-the-fly augmentation
  • Position in the MaldiSuite ecosystem
• 05 - Uncertainty quantification
  • 1. Load the demo dataset
  • 2. Train / calibration / test split
  • 3. Fit the base classifier
  • 4. Monte Carlo Dropout
  • 5. Split conformal prediction
  • 6. Laplace approximation
  • 7. Selective prediction with the uncertainty signal
  • Recap

Example Workflows

Single-Architecture Baseline

import numpy as np
from sklearn.model_selection import StratifiedKFold, cross_val_score
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from maldideepkit import MaldiMLPClassifier

rng = np.random.default_rng(0)
X = rng.standard_normal((400, 6000)).astype("float32")
y = rng.integers(0, 2, size=400)

pipe = Pipeline([
    ("scaler", StandardScaler()),
    ("clf", MaldiMLPClassifier(random_state=0)),
])

cv = StratifiedKFold(n_splits=5, shuffle=True, random_state=0)
scores = cross_val_score(pipe, X, y, cv=cv, scoring="accuracy")
print(f"CV accuracy: {scores.mean():.3f} ± {scores.std():.3f}")

Cross-Architecture Comparison

from maldideepkit import (
    MaldiCNNClassifier,
    MaldiMLPClassifier,
    MaldiResNetClassifier,
    MaldiTransformerClassifier,
)

classifiers = {
    "mlp":         MaldiMLPClassifier(random_state=0),
    "cnn":         MaldiCNNClassifier(random_state=0),
    "resnet":      MaldiResNetClassifier(random_state=0),
    "transformer": MaldiTransformerClassifier(random_state=0),
}

for name, clf in classifiers.items():
    clf.fit(X, y)
    print(f"{name}: {clf.score(X, y):.3f}")

Full MaldiSuite Pipeline

from maldiamrkit import MaldiSet
from maldibatchkit import SpeciesAwareComBat
from maldibatchkit.integrations import MaldiSetAdapter
from maldideepkit import MaldiCNNClassifier

ds = MaldiSet.from_directory(
    "spectra/", "metadata.csv",
    aggregate_by={"antibiotics": "Ciprofloxacin"},
    n_jobs=-1,
)

# Optional batch-effect correction across acquisition sites
adapter = MaldiSetAdapter(batch_column="Batch", species_column="Species")
ds = adapter.correct(ds, SpeciesAwareComBat)

clf = MaldiCNNClassifier(random_state=0).fit(ds, ds.y.squeeze())
clf.save("ciprofloxacin_model")
preds = clf.predict(ds)