ProMS

ProMS selects protein biomarkers from proteomics datasets by integrating additional omics data and applying weighted k-medoids clustering to identify coexpressed protein clusters and representative markers for biomarker discovery and validation.

Key Features:

• Multiview Feature Selection: Selects important features (protein biomarkers) from a target view (e.g., proteomics data) by leveraging information from additional omics datasets to mitigate limitations of small sample sizes in discovery studies.
• Weighted k-Medoids Clustering: Employs a weighted k-medoids clustering algorithm to identify coexpressed protein clusters and select representative proteins as biomarkers.
• Multiomics Extension (ProMS_mo): Extends the approach to multiomics via a constrained weighted k-medoids clustering algorithm (ProMS_mo) to improve marker performance on independent test datasets.
• Functional Interpretation and Flexibility: Provides functional interpretation of selected protein markers through feature clustering, enabling selection of replacement biomarkers for discovery-to-validation transitions.

Scientific Applications:

• Clinical classification: Demonstrated improved performance in clinically important classification problems compared to existing feature selection methods.
• Untargeted mass spectrometry-based proteomics: Applied in untargeted mass spectrometry-based proteomics for protein biomarker discovery with multiomics integration.
• Biomarker panel selection and validation: Supports selection of robust and interpretable protein panels that are more likely to validate across different platforms.

Methodology:

Combines proteomics with other omics datasets (data integration), applies weighted k-medoids clustering to identify coexpressed protein groups and representative markers, and extends to multiomics via a constrained weighted k-medoids algorithm (ProMS_mo).