miQC

miQC models joint quality-control metrics to identify low-quality cells in single-cell RNA-sequencing (scRNA-seq) datasets and improve the reliability of downstream analyses.

Key Features:

• Adaptive Probabilistic Framework: Employs a probabilistic framework to model QC metrics jointly and derive data-driven assessments of cell quality.
• Joint Modeling of QC Metrics: Integrates the proportion of reads mapping to mitochondrial DNA (mtDNA) and the number of detected genes per cell using mixture models to predict low-quality cells and set adaptive thresholds.
• Flexibility Across Datasets: Adapts model parameters across diverse single-cell datasets and experimental conditions, including lower-quality and archived tumor tissue samples.
• Preservation of High-Quality Cells: Minimizes removal of high-quality cells by using data-driven classification rather than fixed cutoffs.

Scientific Applications:

• QC preprocessing for scRNA-seq datasets: Provides adaptive filtering of low-quality cells to improve the fidelity of downstream analyses.
• Quality control of challenging samples: Enables QC in lower-quality tissues and archived tumor samples where fixed cutoffs are unreliable.

Methodology:

Uses mixture models within a probabilistic framework to jointly assess mtDNA read proportion and gene detection count per cell for probabilistic classification of low-quality cells.