Ex
ExIR Feature Classifier
Experimental data-based Integrative Ranking

ExIR analyzes feature behavior within inferred association networks to classify and rank molecular features into four distinct biological roles — without requiring any external annotations or prior knowledge.

🔬 Simulated Association Network
Driver
Biomarker
DE-Mediator
nonDE-Mediator
Unclassified
🎯
Drivers
Highest network impact

Features with the highest impact on biological process or disease progression. They have high centrality in the association network and significant differential expression.

📊
Biomarkers
Highest sensitivity to conditions

Features with the highest sensitivity to different conditions and severity. They show strong differential expression but may have lower network centrality.

🔗
DE-Mediators
Differentially expressed connectors

Features that are differentially expressed in a fluctuating manner, playing mediatory roles between drivers in the association network.

nonDE-Mediators
Non-differentially expressed connectors

Features that are not differentially expressed but have associations with and play mediatory roles between drivers in the network.

🔄 ExIR Workflow
1
Correlation Analysis
Compute Spearman correlations & mutual rank (MR) between all feature pairs using fcor()
2
Network Reconstruction
Build association network from significant correlations using MR-based thresholding
3
Centrality Calculation
Compute IVI (Integrated Value of Influence) combining local, semi-local, and global centrality measures
4
Multi-level Filtration & Scoring
Apply supervised and unsupervised analyses: differential expression, regression, significance testing
5
Classification & Ranking
Integrative ranking outputs: Driver table, Biomarker table, DE-mediator table, nonDE-mediator table
📄 Salavaty A et al. ExIR enables prioritizing driver and biomarker genes from omics data in a reference free manner. iScience. 2026.
💻 Available in R/Python via the influential package