SPsimSeq

SPsimSeq generates simulated bulk and single-cell RNA sequencing (RNA-seq) data that reproduce gene expression marginal distributions and inter-gene dependencies for benchmarking and method evaluation.

Key Features:

• Semi-parametric density estimation: SPsimSeq uses a specially designed exponential family to estimate gene expression distributions from real bulk or single-cell RNA-seq data.
• Gaussian-copula dependence modeling: It employs Gaussian-copulas to simulate datasets from the estimated marginals while preserving inter-gene dependence structure.
• Modeling biological signals and confounders: The method can incorporate differential expression signals and account for confounding batch effects across experimental scenarios.
• Flexible experimental design simulation: It supports simulation of multiple groups and batches with specified sample sizes and library sizes.
• Implementation: SPsimSeq is implemented as an R package.

Scientific Applications:

• Benchmarking and method evaluation: Generating realistic RNA-seq datasets for validating bioinformatics pipelines and benchmarking analysis methods.
• Power analysis: Enabling power analyses by simulating datasets with controlled sample sizes, library sizes, and effect sizes.
• Method development and testing: Testing statistical models and algorithms under realistic dependence structures and marginal distributions.
• Batch effect assessment: Studying the impact of batch effects and other confounding factors on gene expression analyses.

Methodology:

SPsimSeq constructs marginal gene expression distributions from real RNA-seq data via a specially designed exponential-family density estimator and then generates simulated datasets by sampling from these marginals using Gaussian-copulas to preserve inter-gene dependencies.