OmniDoublet: a method for doublet detection in multimodal single-cell sequencing data.

Area of Science:

• Single-cell genomics
• Computational biology
• Bioinformatics

Background:

• Doublets in single-cell sequencing data introduce biases, compromising downstream analyses.
• Current doublet detection methods are often limited to single-modality data and lack robustness across datasets.

Purpose of the Study:

• To develop a robust multimodal doublet detection method integrating transcriptomic and epigenomic data.
• To overcome the limitations of existing single-modality doublet detection approaches.

Main Methods:

• Developed OmniDoublet, a multimodal method integrating transcriptomic and epigenomic data.
• Utilized Jaccard similarity for cross-modality cell reliability weighting.
• Combined modality-specific doublet scores into an integrated score.
• Employed a Gaussian mixture model (GMM) for thresholding and binary classification.

Main Results:

• OmniDoublet demonstrated superior accuracy, robustness, and scalability in benchmarking against state-of-the-art methods.
• The method effectively integrates multimodal data for enhanced doublet detection.
• Accurate binary classification of cells as singlets or doublets was achieved.

Conclusions:

• OmniDoublet provides a robust framework for doublet detection across diverse single-cell sequencing scenarios.
• Harnessing multimodal data significantly enhances doublet detection accuracy and reliability.
• The method enables more accurate insights into cellular processes by mitigating doublet-induced biases.