visTrajectory

visTrajectory extracts and visualizes latent one-dimensional trajectories and their associated uncertainty in high-dimensional multivariate biological datasets using Bayesian Unidimensional Scaling (BUDS).

Key Features:

• Bayesian Unidimensional Scaling (BUDS): maps multivariate observations onto latent one-dimensional coordinates that represent an underlying trajectory.
• Estimated uncertainty bounds: provides posterior-based uncertainty estimates for trajectory positions expressed as confidence contours for individual data points.
• Statistical modeling of dissimilarities and posterior sampling: models pairwise dissimilarities and generates posterior samples for trajectory estimation.
• DiSTATIS registration: applies DiSTATIS registration methods to posterior samples to align and summarize trajectory estimates.
• Density clouds: generates density clouds to illustrate data density and variability across regions of the trajectory.
• Automatic visualization generation: produces visual summaries of trajectories and uncertainty for exploratory analysis.

Scientific Applications:

• Microbial community composition (16S): recovers and visualizes natural orderings in microbiome 16S datasets to aid interpretation of microbial dynamics.
• Gene expression (RNA-seq): reveals latent continuums and dominant sources of variation in RNA-seq gene expression datasets.
• Roll call data analysis: identifies and visualizes latent orderings and uncertainties in roll call datasets.

Methodology:

Performs Bayesian Unidimensional Scaling (BUDS) with statistical modeling of dissimilarities and posterior sampling, applies DiSTATIS registration to posterior samples, and visualizes results via confidence contours and density clouds along a latent one-dimensional trajectory.