scTsI: an effective two-stage imputation method for single-cell RNA-seq data.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for understanding cellular heterogeneity in development and disease.
• Dropout events, caused by technical noise or low sequencing depth, significantly impair scRNA-seq data analyses.
• Existing imputation methods may introduce noise or alter expression values, limiting their effectiveness, especially with high dropout rates.

Purpose of the Study:

• To develop an advanced imputation algorithm, scTsI, to accurately address dropout events in scRNA-seq data.
• To preserve original high expression values and avoid introducing new noise during imputation.
• To improve the performance of various downstream analyses, including visualization, clustering, and trajectory inference.

Main Methods:

• scTsI employs a two-stage imputation process.
• Stage one imputes zero values using information from neighboring cells and genes.
• Stage two transforms the expression matrix, adjusts imputed values via ridge regression, and uses bulk RNA-seq data as a constraint, while preserving original high expression values and allowing sparse matrix input.

Main Results:

• scTsI successfully restores gene expression levels in scRNA-seq data across various dropout rates and data dimensions.
• The algorithm maintains cell-cell similarity, outperforming existing imputation methods.
• scTsI demonstrably improves the accuracy of data visualization, clustering, and cell trajectory inference.

Conclusions:

• scTsI offers a robust solution for handling dropout events in scRNA-seq data.
• The method accurately imputes missing values, preserves biological information, and enhances downstream analytical outcomes.
• scTsI is a valuable tool for researchers working with scRNA-seq data, particularly in complex biological systems and disease studies.