GTestimate

GTestimate applies the Simple Good-Turing estimator to normalize single-cell RNA sequencing (scRNA-seq) data by estimating per-cell relative gene expression while accounting for unobserved genes.

Key Features:

• Simple Good-Turing estimator: Leverages the Simple Good-Turing estimator to refine per-cell relative gene expression estimates by incorporating information about unobserved genes.
• Improved accuracy of expression and distances: Enhances relative gene expression and cell–cell distance estimations to benefit downstream clustering, differential expression, and trajectory inference.
• Validation with cta-seq: Performance validated using cell-targeted PCR-amplification (cta-seq) with ultra-deep sequencing of individual cells.
• Compatibility with Seurat workflows: Designed for compatibility with analyses that use Seurat for scRNA-seq downstream processing.
• Tested on multiple datasets: Evaluated on three example datasets to assess effects on downstream analyses.

Scientific Applications:

• Normalization for scRNA-seq: Produces more accurate per-cell normalization in scRNA-seq studies, particularly under shallow sequencing depths.
• Downstream analysis improvement: Improves clustering, differential expression analysis, and trajectory inference by providing more accurate expression and distance estimates.
• Studies of cellular heterogeneity and regulatory networks: Supports analyses of cellular heterogeneity, gene regulatory networks, and developmental processes by accounting for unobserved genes.

Methodology:

Applies the Simple Good-Turing estimator to scRNA-seq count data to estimate per-cell relative gene expression, replacing maximum-likelihood estimators and accounting for unobserved genes.