netDx

netDx is a machine learning tool for patient classification using multi-modal data, including clinical and genomic information. It aims to support precision medicine by providing an interpretable and robust method for predicting patient outcomes, such as binary cancer survival.

Key features of netDx include:

1. Network-based approach: netDx converts patient data into networks of patient similarity, which aligns with the intuitive approach used by clinicians for medical diagnosis.

2. Feature selection: The tool performs feature selection to identify the most informative features for classification.

3. Data integration: netDx can integrate features from multiple data types, enhancing the classifier's performance.

4. Missing data handling: The tool can handle missing data without requiring imputation, which is a common issue in real-world datasets.

5. Interpretability: netDx provides native support for grouping genes into pathways, offering mechanistic insights into predictive features.

The netDx Bioconductor package offers various workflows for users to build custom patient classifiers, including turnkey functions for one-step predictor generation from multi-modal data and parallel execution for improved speed. Users can compute model performance metrics, such as AUROC, AUPR, and accuracy, using built-in functions and examples. The package also utilizes RCy3 to visualize top-scoring pathways and the final integrated patient network in Cytoscape.