netDx: Software for building interpretable patient classifiers by multi-'omic data integration using patient...

Area of Science:

• Computational biology
• Bioinformatics
• Machine learning in healthcare

Background:

• Precision medicine relies on patient classification using clinical and genomic data.
• Genomic data analysis faces challenges like small sample sizes and overfitting risk.
• Interpretable models are crucial for understanding genomic data in clinical settings.

Purpose of the Study:

• To introduce netDx, a machine learning method for integrating multi-modal patient data to build patient classifiers.
• To enhance patient classification through network-based similarity, mirroring clinical diagnostic approaches.
• To provide a user-friendly Bioconductor package for custom patient classifier development.

Main Methods:

• netDx converts patient data into intuitive networks of patient similarity.
• It integrates selected features and assigns new patients based on profile similarity.
• The method handles missing data without imputation and supports pathway grouping for mechanistic insights.

Main Results:

• netDx demonstrates excellent performance, outperforming many machine learning methods in binary cancer survival prediction.
• The method effectively handles missing data, a common issue in real-world datasets.
• netDx provides excellent interpretability by grouping genes into pathways for mechanistic insights.

Conclusions:

• netDx offers a robust and interpretable approach to patient classification using multi-modal data.
• The netDx Bioconductor package facilitates custom classifier development with versatile workflows and performance metrics.
• This method advances precision medicine by enabling pathway-based classification from sparse genetic data.