miQC

miQC jointly models the proportion of reads mapping to mitochondrial DNA encoded genes and the number of detected genes with mixture models in a probabilistic framework to predict low-quality cells in single-cell RNA-sequencing datasets.

Key Features:

• Joint modeling of QC metrics: miQC simultaneously analyzes both mitochondrial DNA read proportion and number of detected genes rather than applying independent thresholds.
• Adaptive probabilistic framework: The method uses mixture models to adapt to different types of single-cell datasets without requiring arbitrary or species-specific threshold settings.
• Data-driven quality prediction: miQC calculates the probability that each cell is low-quality based on its specific combination of QC metrics.

Scientific Applications:

• Quality control of scRNA-seq data: miQC identifies and removes cells with poor or degraded sample quality prior to downstream analyses.
• Analysis of challenging tissue types: miQC preserves high-quality cells from archived tumor tissues and tissues associated with mitochondrial function, such as kidney tissue, where standard QC thresholds are overly stringent.

Methodology:

miQC employs mixture models within a probabilistic framework to jointly model the proportion of reads mapping to mitochondrial DNA encoded genes and the number of detected genes.