censcyt

censcyt performs differential abundance analysis of high-dimensional cytometry data by modeling censored covariates, such as right-censored survival times, as predictors to detect associations between cell populations and censored clinical variables.

Key Features:

• Extension of diffcyt: Extends the diffcyt workflow including preprocessing, cell population identification, and differential testing for associations with binary or continuous covariates, enabling inclusion of censored covariates.
• Handling Censored Covariates: Implements reversed association testing by modeling censored variables (e.g., survival time) as predictors rather than responses to accommodate right-censoring in covariates.
• Generalized Linear Mixed Models (GLMMs): Incorporates censored covariates directly into GLMMs as an alternative to Cox proportional hazards approaches, providing controlled error rates with reasonable sensitivity.
• Simulation and Case Studies: Validated through simulation studies and a case study demonstrating capability to manage censored data in cytometry datasets.
• Implementation: Implemented in R.

Scientific Applications:

• Clinical Research: Enables differential abundance analysis in clinical studies where survival time is right-censored, preserving information from censored covariates.
• Single Cell Analysis: Supports analysis of cell composition changes and state transitions in single-cell cytometry datasets when relevant covariates are censored.

Methodology:

Builds on diffcyt preprocessing and cell population identification, performs differential testing for associations with binary or continuous covariates, applies reversed-association modeling of censored covariates as predictors, fits GLMMs that incorporate censored predictors, and is validated with simulation studies and a case study.