FXI Deficiency Classifier

This repository contains code and analysis for predicting bleeding phenotype in Factor XI (FXI) deficiency.

Title: Plasma proteomic signatures determine clot stability and predict bleeding risk in factor XI deficiency

Authors: Stéphanie E. Reitsma, Tracey G. Oellerich, Olivia C. White, Michael L. Hill, Daniel Kissau, John Burthem, Dougald M. Monroe, Paula H. B. Bolton-Maggs, Paul S. de Vries, Suzanne S. Sindi, Karin Leiderman, Gillian N. Gidley, Kirk C. Hansen, Alisa S. Wolberg

Overview

FXI_Deficiency_Classifier.ipynb allows users to build and evaluate logistic regression models to predict FXI deficiency. Key features include:

• User-defined inputs:

  • Fixed features to include in every model
  • Number of features per model
  • Output file name

• Model evaluation:

  • Systematically evaluates all feature combinations using 5-fold cross-validation
  • Computes performance metrics:
    • Mean AUROC with 95% confidence intervals
    • Optimal classification thresholds (Youden’s J statistic)
    • Confusion matrices
    • Standard rates: TPR, FNR, FPR, TNR

• Outputs:

  • Excel file containing results
  • Top 5 models sorted by mean AUROC
  • ROC curve plots for Manchester and Indeterminate datasets

• Data files included:

  • Discovery data: FXI_Deficient_Bleeders_NonBleeders_data.xlsx
  • Hold-out data: Indeterminant_data.xlsx

Installation

Clone the repository and install dependencies:

git clone https://github.com/karinleiderman/LeidermanResearchGroupCode/tree/main/FXI_Deficiency_Classifier.git
cd FXI_Deficiency_Classifier
pip install pandas scikit-learn matplotlib numpy scipy

Contact

If you have any questions or GitHub issues, feel free to contact us through email (toelleri@unc.edu & karin.leiderman@unc.edu).