Semi-parametric empirical bayes method for multiplet detection in snATAC-seq with probabilistic multi-omic...

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Multiplets, arising from multiple cells in single-cell sequencing droplets, create distorted molecular profiles.
• Detecting multiplets in single-nucleus ATAC-seq (snATAC-seq) data is difficult due to sparse and overdispersed chromatin accessibility data.
• Computational methods integrating multi-feature and multi-modal evidence are needed for accurate multiplet detection.

Purpose of the Study:

• Introduce SEBULA, a novel semi-parametric empirical Bayes framework for multiplet detection in snATAC-seq data.
• Develop a method that models the singlet background directly from observed chromatin accessibility signals, avoiding synthetic doublets.
• Extend SEBULA to integrate complementary evidence from additional features and modalities, such as gene expression.

Main Methods:

• SEBULA models the singlet background using fragment-level chromatin accessibility signals from snATAC-seq data.
• The framework generates classification probabilities for direct false discovery rate control.
• SEBULA integrates evidence from multiple features and modalities, including gene expression profiles.

Main Results:

• SEBULA demonstrates improved sensitivity and specificity compared to existing snATAC-seq methods across simulations and seven multimodal datasets.
• The evidence integration framework achieves performance comparable or superior to state-of-the-art multiomic approaches.
• SEBULA maintains computational efficiency while providing robust multiplet detection.

Conclusions:

• SEBULA offers a powerful and accurate approach for multiplet detection in snATAC-seq data.
• The framework's ability to integrate multiomic data enhances multiplet identification accuracy.
• SEBULA provides a reliable tool for improving downstream analyses in single-cell genomics.