SampleQC

SampleQC performs robust multivariate quality control for single-cell RNA sequencing (scRNA-seq) datasets by fitting Gaussian mixture models across multiple samples to separate technical artifacts from biological variability and preserve rare cell types.

Key Features:

• Gaussian mixture model (GMM)-based multivariate QC: fits GMMs across multiple samples to model joint distributions of quality metrics.
• Multivariate integration of QC metrics: integrates library size, number of features observed, and mitochondrial proportion in a multivariate framework rather than relying on univariate thresholds.
• Sensitivity to rare cell types: reduces the exclusion of rare cell populations compared to traditional univariate QC methods.
• Multi-sample modeling: models data across multiple samples to account for sample-level variation.
• Scalability: demonstrated capability on a dataset of 867,000 cells across 172 samples.
• Implementation: provided as an R package.
• Validation: performance shown using simulations and real-world datasets.
• Distinction of technical artifacts and biological variability: uses multivariate GMM fitting to differentiate technical artifacts from true biological variation.

Scientific Applications:

• Multi-sample, multi-cell-type scRNA-seq QC: applies to quality control in complex scRNA-seq studies containing multiple cell types and samples.
• Retention of rare cell populations: improves detection and retention of rare cell types during QC procedures.
• Large-scale scRNA-seq studies: applicable to studies comprising hundreds of samples and hundreds of thousands of cells.
• High-dimensional biological data QC: generalizable to other high-dimensional biological datasets where multivariate QC is required.

Methodology:

SampleQC fits Gaussian mixture models (GMMs) across samples using multivariate quality metrics (library size, number of features observed, mitochondrial proportion) and was validated by simulations and application to a dataset of 867,000 cells across 172 samples.