PCLassoReg

PCLassoReg identifies risk protein complexes in cancer by applying a protein complex–based group Lasso-logistic regularized learning framework to proteomic data.

Key Features:

• Protein Complex-Based Framework: Integrates protein complexes into a regularized learning framework using a group Lasso-logistic model (PCLassoLog) to select complex-level predictors.
• Superior Predictive Performance: Experimental validation using deep proteomic data from two cancer types showed PCLassoLog outperformed traditional methods on independent datasets.
• Comprehensive Risk Identification: Identifies risk protein complexes that include individual risk proteins and their interacting partners contributing synergistically to cancer progression.
• Selection Probabilities Calculation: Computes selection probabilities to provide a probabilistic measure of each complex's relevance.
• Complementary Models: Includes PCLassoLog alongside two additional protein complex-based models to complement complex identification.
• Pan-Cancer Analysis Capability: Facilitates identification of risk protein complexes across multiple cancers, demonstrated across 12 cancer types.
• Association with Gene Mutation: Detects associations between identified protein complexes and gene mutations.

Scientific Applications:

• Cancer Classification: Classifies cancer samples based on complex-level predictive features derived from PCLassoLog.
• Risk Protein Complex Discovery: Discovers protein complexes comprising risk proteins and synergistic partners that may drive cancer mechanisms.
• Genetic Research: Associates protein complexes with gene mutations to support studies of molecular mechanisms underlying cancer.

Methodology:

Uses a group Lasso-logistic regression model (PCLassoLog) within a regularized learning framework, integrates and analyzes deep proteomic data, computes selection probabilities, and employs two additional protein complex-based models for complementary selection.