baredSC

baredSC applies Bayesian inference to estimate intrinsic gene expression distributions from single-cell RNA sequencing (scRNA-seq) data while mitigating Poisson sampling noise and data sparsity.

Key Features:

• Bayesian inference: Employs a Gaussian mixture model within a Bayesian framework to infer intrinsic expression distributions and mitigate Poisson sampling errors.
• Dimensionality flexibility: Performs one-dimensional analysis for individual genes and two-dimensional analysis to estimate correlations between gene pairs.
• Handling sparsity: Recovers underlying expression signals in sparse scRNA-seq data that are obscured by technical sampling noise.
• Multi-modal distribution recovery: Effectively detects multi-modal expression distributions that traditional methods may miss in noisy single-cell data.

Scientific Applications:

• Gene expression distribution analysis: Estimates probability density functions (PDFs) of gene expression to reveal intrinsic variability within single-cell datasets.
• Correlation estimation: Identifies correlations between gene pairs to inform studies of genetic interactions and regulatory networks.
• Simulated data analysis: Recovers complex multi-modal expression distributions in simulated scRNA-seq datasets.
• Real biological datasets: Applied to embryonic limb data to measure an anti-correlation between Hoxd13 and Hoxa11 and to embryonic hindlimb data to detect a trimodal distribution of Pitx1.

Methodology:

Uses a Gaussian mixture model within a Bayesian inference framework to infer one- and two-dimensional intrinsic expression distributions while accounting for Poisson sampling noise.