Synthetic DNA barcodes identify singlets in scRNA-seq datasets and evaluate doublet algorithms

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) is a powerful technique for analyzing cellular heterogeneity.
• scRNA-seq data often contain doublets, which are pairs of cells captured as a single event, confounding downstream analysis.
• Existing doublet detection methods lack robust evaluation due to the absence of ground-truth data.

Purpose of the Study:

• To develop a framework for generating ground-truth singlets in scRNA-seq data.
• To provide a reliable method for evaluating existing doublet detection algorithms.
• To improve the accuracy of doublet detection in scRNA-seq analysis.

Main Methods:

• Leveraged scRNA-seq datasets with synthetically introduced DNA barcodes.
• Developed the 'singletCode' framework to extract ground-truth singlets.
• Utilized ground-truth singlets to train and evaluate a machine learning classifier for doublet detection.

Main Results:

• Demonstrated the feasibility of the singletCode framework for generating ground-truth singlets.
• Successfully evaluated the performance of existing doublet detection methods.
• Developed a proof-of-concept machine learning classifier that outperformed other algorithms.

Conclusions:

• The singletCode framework provides a robust method for identifying ground-truth singlets.
• This framework enables reliable evaluation and improvement of doublet detection strategies.
• The developed approach enhances the accuracy of scRNA-seq data analysis by enabling robust doublet detection.