COMPASS

COMPASS employs a Bayesian hierarchical model to analyze high-dimensional single-cell immunological data (e.g., flow cytometry) and quantify antigen-specific, polyfunctional T-cell subset responses for correlation with clinical outcomes.

Key Features:

• Bayesian Hierarchical Framework: Uses a Bayesian hierarchical model to jointly model observed cell subsets and assess antigen-specific responses.
• Posterior Probability Quantification: Computes posterior probabilities for each cell subset to quantify the likelihood of antigen specificity at the subset and subject levels.
• Summary Statistics for Subject-Level Analysis: Derives two summary statistics that capture polyfunctional response quality in individual subjects for downstream association with clinical outcomes.
• Modeling of All Observed Cell Subsets: Models all observed combinatorial cell subsets to identify those most likely to exhibit antigen-specific responses.
• Regularization of Small Cell Counts: Regularizes small cell counts to stabilize inference across rare subsets.
• Identification of Cellular Correlates of Protection/Immunity: Has revealed cellular correlates of protection and immunity in the RV144 HIV vaccine efficacy trial that were not detected by other methods.

Scientific Applications:

• Immune response profiling: Analyze antigen-specific T-cell subset responses to infections or vaccinations using high-dimensional single-cell data.
• Vaccine efficacy and correlates analysis: Identify and quantify cellular correlates of protection and relate polyfunctional responses to clinical outcomes, as demonstrated in RV144.
• Clinical research and biomarker discovery: Generate subject-level metrics for association studies linking immune response quality to clinical endpoints.

Methodology:

Applies a Bayesian hierarchical model to high-dimensional single-cell (e.g., flow cytometry) data by modeling all observed cell subsets, regularizing small cell counts, computing posterior probabilities per subset, and deriving two subject-level summary statistics.