Probability of stealth multiplets in sample-multiplexing for droplet-based single-cell analysis.

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Droplet-based single-cell RNA sequencing (scRNA-seq) is limited by multiplets (multiple cells per droplet).
• Sample-multiplexing scRNA-seq (mx-scRNA-seq) reduces multiplets but can still contain undetectable 'stealth multiplets'.

Purpose of the Study:

• To quantitatively examine the probability and impact of stealth multiplets in mx-scRNA-seq.
• To introduce a theoretical model for predicting different classes of multiplets in mx-scRNA-seq.

Main Methods:

• Development of a theoretical model to predict four classes of multiplets: homogeneous stealth, partial stealth, multilabelled, and unlabelled.
• Estimation of the probability of each multiplet class.
• Demonstration of stealth multiplet presence in real mx-scRNA-seq datasets using oligonucleotide-barcode and genotype-based demultiplexing.

Main Results:

• Partial stealth multiplets, previously overlooked, can significantly impact mx-scRNA-seq datasets, especially with suboptimal labeling or demultiplexing.
• The theoretical model successfully predicted and quantified various multiplet classes.
• Stealth multiplets were confirmed in both oligonucleotide-barcode and genotype-based demultiplexing strategies.

Conclusions:

• Optimizing labeling procedures and selecting appropriate demultiplexing algorithms are crucial for mx-scRNA-seq integrity.
• The developed theoretical framework provides a basis for estimating multiplet probabilities.
• Ensuring data integrity in mx-scRNA-seq requires careful consideration of potential stealth multiplets.