LogMiNeR

LogMiNeR applies network-constrained logistic regression to integrate prior biological knowledge into predictive, interpretable models of high-dimensional transcriptional profiling data for systems immunology.

Key Features:

• Network-constrained logistic regression: Integrates network-level constraints into logistic regression models using transcriptional profiling data to incorporate prior biological knowledge.
• Multiple network constraints: Supports the use of multiple networks that encode different types of prior knowledge.
• Improved interpretability: Aligns model coefficients with known pathways and networks to produce biologically meaningful predictors.
• Addresses limitations of standard methods: Mitigates generation of biologically uninterpretable predictors and multiple equally predictive but narrow models common in traditional statistical learning.
• Applied to vaccination studies: Has been used to analyze influenza vaccination responses and identified B cell-specific genes and mTOR signaling associated with effective responses in young adults.

Scientific Applications:

• Systems immunology: Derives interpretable predictive models from high-dimensional immune profiling data.
• Vaccination response analysis: Analyzes differential responses to influenza vaccination to identify cell-type-specific genes and signaling pathways such as B cells and mTOR.
• Infection and disease dysregulation profiling: Applies to large-scale transcriptional profiling experiments related to infection, vaccination, and disease dysregulation.

Methodology:

LogMiNeR fits logistic regression models constrained by biological networks, incorporating prior knowledge such as gene expression linked to specific pathways or cellular processes and supporting multiple network constraints.