Confound-Corrected Connectome-Based Predictive Modelling in Python

Confound-Corrected Connectome-Based Predictive Modelling is a Python package for performing connectome-based predictive modeling (CPM). This toolbox is designed for researchers in neuroscience and psychiatry, providing robust methods for building predictive models based on structural or functional connectome data. It emphasizes replicability, interpretability, and flexibility, making it a valuable tool for analyzing brain connectivity and its relationship to behavior or clinical outcomes.

What is Connectome-Based Predictive Modeling?

Connectome-based predictive modeling (CPM) is a machine learning framework that leverages the brain's connectivity patterns to predict individual differences in behavior, cognition, or clinical status. By identifying key edges in the connectome, CPM creates models that link connectivity metrics with target variables (e.g., clinical scores). This approach is particularly suited for studying complex relationships in neuroimaging data and developing interpretable predictive models.

Key Features

• Univariate Edge Selection: Supports methods like pearson, spearman, and their partial correlation counterparts, with options for p-threshold optimization and FDR correction.
• Cross-Validation: Implements nested cross-validation for robust model evaluation.
• Edge Stability: Selects stable edges across folds to improve model reliability.
• Confound Adjustment: Controls for covariates during edge selection and modeling.
• Permutation Testing: Assesses the statistical significance of models using robust permutation-based methods.

Documentation

For detailed instructions on installation, usage, and advanced configurations, visit the documentation website.

Installation

Install the latest release from PyPI:

pip install cccpm

Or install the development version from GitHub:

git clone https://github.com/wwu-mmll/confound_corrected_cpm.git
cd confound_corrected_cpm
pip install .

CCCPM requires Python 3.10–3.14 and uses PyTorch for computation (CPU by default; CUDA/MPS used automatically when available). See the installation guide for platform-specific notes.

Quick Example

Here's a quick overview of how to run a CPM analysis:

from cccpm import CPMAnalysis, UnivariateEdgeSelection, PThreshold
from sklearn.model_selection import KFold

# Configure edge selection. `selection_input="residualized"` controls for the
# covariates while *selecting* edges.
univariate_edge_selection = UnivariateEdgeSelection(
    selection_statistic="pearson",
    selection_input="residualized",
    edge_selection=[PThreshold(threshold=[0.05], correction=["fdr_by"])]
)

# Create the CPM analysis object. `model_input` is the second, independent
# confound-control choice: deconfound the connectome the models consume.
cpm = CPMAnalysis(
    results_directory="results/",
    cv=KFold(n_splits=10, shuffle=True, random_state=42),
    edge_selection=univariate_edge_selection,
    model_input="residualized",
    n_permutations=100
)

# Run the analysis
X = ...           # Connectome data, shape (n_samples, n_features)
y = ...           # Target variable, shape (n_samples,)
covariates = ...  # Covariates to control for, shape (n_samples, n_covariates)
cpm.run(X=X, y=y, covariates=covariates)   # covariates are optional

Contributing

Contributions are welcome! If you have ideas, feedback, or feature requests, feel free to open an issue or submit a pull request on the GitHub repository.