Propensity score matching enables batch-effect-corrected imputation in single-cell RNA-seq analysis.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) offers high-resolution biological insights.
• Large scRNA-seq datasets are often compromised by technical noise, including batch effects and dropouts.
• Existing methods struggle to simultaneously address batch effects and dropout imputation effectively.

Purpose of the Study:

• To develop a novel method for integrated batch effect correction and dropout imputation in scRNA-seq data.
• To improve the accuracy and reliability of scRNA-seq data analysis.
• To enhance the recovery of biologically meaningful expression patterns.

Main Methods:

• Proposed a novel propensity score matching method for scRNA-seq data (scPSM).
• Leveraged causal inference principles by borrowing information from similar cells in deep sequenced batches.
• Employed a weighted averaging approach to integrate information across similar cells.

Main Results:

• scPSM demonstrated superior performance compared to state-of-the-art methods on simulation and real datasets.
• Improved clustering accuracy by separating cell types while correcting for batch effects.
• Enabled batch and dropout-free differential expression analysis.
• Achieved effective denoising while preserving genuine biological structures.
• Showed robustness to hyperparameters and small datasets.

Conclusions:

• scPSM offers a powerful, integrated solution for scRNA-seq data challenges.
• The method effectively corrects batch effects, imputes dropouts, and denoises data simultaneously.
• scPSM enhances downstream analyses, leading to more accurate biological interpretations.