Imputing single-cell RNA-seq data by considering cell heterogeneity and prior expression of dropouts

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for understanding cellular expression patterns.
• scRNA-seq data analysis faces challenges from technical noise, particularly dropout events (unobserved gene expression).
• Dropout events create artificial zeros, complicating accurate gene expression interpretation.

Purpose of the Study:

• To develop a novel computational method for imputing dropout events in scRNA-seq data.
• To address cell heterogeneity and the relationship between dropout rates and expression levels.
• To enhance the accuracy and robustness of scRNA-seq data analysis.

Main Methods:

• A cell sub-population based bounded low-rank (PBLR) method was developed.
• The method accounts for cell heterogeneity and dropout rate-expression level relationships.
• PBLR was applied to both simulated and real scRNA-seq datasets.

Main Results:

• PBLR effectively recovers dropout events in scRNA-seq data.
• The method significantly improves low-dimensional data representation.
• PBLR enhances the recovery of gene-gene relationships obscured by dropout events.
• Accurate and robust cell sub-populations were automatically detected by PBLR.

Conclusions:

• PBLR offers a flexible and generalizable approach for scRNA-seq data analysis.
• The method overcomes key limitations posed by dropout events.
• PBLR improves the biological insights obtainable from scRNA-seq experiments.