MitoChontrol: Adaptive mitochondrial filtering for robust single-cell RNA sequencing quality control.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Mitochondrial transcript abundance is a key quality control metric in single-cell RNA sequencing (scRNA-seq).
• Fixed percentage thresholds for mitochondrial content are inadequate due to biological variation across cell types and tissues.
• This can lead to the exclusion of viable cells or retention of compromised cells.

Purpose of the Study:

• To develop a novel, cell-type-aware probabilistic framework for mitochondrial quality control in scRNA-seq.
• To improve the accuracy of identifying compromised cells while preserving biologically relevant cell populations.

Main Methods:

• MitoChontrol models mitochondrial transcript fraction within transcriptionally coherent clusters using a Gaussian mixture distribution.
• Compromised cells are identified from the upper tail of cluster-specific distributions.
• Filtering thresholds are determined by a user-defined confidence value for cellular compromise probability.

Main Results:

• MitoChontrol effectively distinguishes transcriptionally compromised cells from viable cells with biologically elevated mitochondrial content.
• The framework outperforms traditional fixed-threshold and outlier-based methods.
• Application to perturbation experiments and a pancreatic cancer dataset demonstrated its efficacy.

Conclusions:

• MitoChontrol provides a more accurate and nuanced approach to mitochondrial quality control in scRNA-seq.
• This method enhances the reliability of scRNA-seq data by preserving biologically relevant cell populations.
• The tool is freely available and integrates with existing bioinformatics workflows.