PATHcrosstalk

PATHcrosstalk implements crosstalk correction as an R package to refine pathway deregulation scores (PDS) computed from gene expression using the Pathifier algorithm and thereby improve prognostic analyses in cancer, including hepatocellular carcinoma (HCC).

Key Features:

• Crosstalk Correction: Introduces a procedure to account for gene overlap between pathway pairs, refining estimation of pathway deregulation scores (PDS).
• Integration with Pathifier Algorithm: Builds on the Pathifier algorithm to compute PDS from gene expression data.
• Disease-Specific Feature Identification: Identifies disease-specific pathway features and governing genes that predict prognosis in hepatocellular carcinoma (HCC).

Scientific Applications:

• Prognostic Analysis in Cancer Research: Improves robustness and accuracy of prognostic predictions compared with single-gene (SG) lists on independent datasets.
• Enhanced Predictive Accuracy: In The Cancer Genome Atlas (TCGA) HCC cohort of 355 patients, cross-validation showed crosstalk-corrected PDSs increased prediction accuracy by an average of 10.2% versus SG features and comparisons including deep learning approaches.
• External Validation: Crosstalk-corrected features consistently outperformed traditional single-gene features on external HCC datasets.
• Insight into Cancer Hallmarks: By identifying governing genes within refined pathway features, provides insights into cancer hallmarks specific to HCC.
• Reproducible Features: Pathway-based integration of biological knowledge yields features reproducible across different datasets.

Methodology:

Computes pathway deregulation scores (PDS) using the Pathifier algorithm on gene expression data and applies a crosstalk-correction procedure to account for gene overlap between pathways; evaluation used cross-validation and comparisons to single-gene (SG) features including deep learning approaches.