VPAC

VPAC performs variational projection-based clustering of single-cell transcriptomic data to identify putative cell types from single-cell RNA-sequencing (scRNA-seq) datasets.

Key Features:

• Model-Based Algorithm: Assumes single-cell samples follow a Gaussian mixture distribution within a latent space to inform clustering.
• Variational Projection: Projects high-dimensional transcriptomic data into a latent space using a variational projection approach to improve separability.
• Scalability and Robustness: Validated to perform across varying data dimensionalities, dataset sizes, and levels of sparsity.
• Input Data Types: Supports both discrete count data and normalized continuous expression data as input.
• Gene Signature Detection: Identifies genes with strong, unique signatures specific to particular cell types.

Scientific Applications:

• Cell-type identification from scRNA-seq: Partitioning cells into putative cell types based on single-cell RNA-sequencing profiles.
• Systems biology: Characterizing cell identity and functionality to inform systems-level biological studies.
• Analysis of complex datasets: Applying to large, sparse, high-dimensional biological datasets and other domains with similar data characteristics.

Methodology:

Applies variational projection to map data into a latent space where a Gaussian mixture distribution is assumed for clustering and accepts discrete count and normalized continuous inputs.