CYBERTRACK

CYBERTRACK models time-series flow cytometry data to cluster and track dynamic cell populations using a time-dependent multivariate Gaussian mixture framework.

Key Features:

• Implementation: Provided as an R package for analysis of time-series flow cytometry datasets.
• Multivariate Gaussian Mixture Assumption: Models flow cytometry observations as arising from a multivariate Gaussian mixture distribution.
• Model-Based Clustering: Performs clustering by fitting the multivariate Gaussian mixture to identify distinct cell populations.
• Time-Dependent Tracking: Links mixture component proportions across successive timepoints so that the proportion at a given time depends on its previous value.
• Change-Point Detection: Detects significant shifts or change-points in the overall mixture proportions across the dataset timeline.
• Parameter Estimation and Validation: Estimates parameters of the mixture model and has been validated on simulation data and two real flow cytometry datasets.

Scientific Applications:

• Cellular behavior and immune response studies: Analysis of dynamic changes in cell populations over time to study cellular behavior and immune responses.
• Lymphocyte population dynamics: Characterization of lymphocyte dynamics, with results reported to align with established knowledge about lymphocyte behavior.
• Time-series flow cytometry analysis: Systematic estimation of time-dependent population transitions and identification of temporal change-points in longitudinal flow cytometry experiments.

Methodology:

Model-based clustering using a multivariate Gaussian mixture model with component proportions that depend on previous timepoints; change-point detection on overall mixture proportions; parameter estimation; validation on simulated data and two real flow cytometry datasets.