scMTD: a statistical multidimensional imputation method for single-cell RNA-seq data leveraging transcriptome dynamic...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) enables high-resolution transcriptome analysis.
• Dropout events in scRNA-seq data introduce noise, distorting gene expression levels and biological insights.
• Inaccurate data hinders reliable downstream analyses, necessitating robust imputation methods.

Purpose of the Study:

• To develop a novel statistical model-based imputation algorithm for scRNA-seq data.
• To address the challenge of dropout events and improve the accuracy of gene expression profiles.
• To enhance the performance of various downstream analyses in scRNA-seq studies.

Main Methods:

• Developed scMTD, a multidimensional imputation algorithm utilizing statistical modeling.
• Incorporated cell-level, gene-level, and transcriptome dynamics information.
• Leveraged pseudo-time ordering to identify local cell neighbors and gene co-expression networks.

Main Results:

• scMTD effectively recovered biological signals in scRNA-seq data, outperforming existing imputation methods.
• Demonstrated significant improvements in Fluorescence In Situ Hybridization (FISH) validation, trajectory inference, and differential expression analysis.
• Enhanced clustering analysis and improved the identification of cell types and rare cell populations.

Conclusions:

• scMTD preserves essential gene expression characteristics and improves cell subpopulation clustering.
• The algorithm aids in studying gene expression dynamics and discovering rare cell types.
• scMTD is reliable, applicable to both UMI-based and non-UMI-based data, and scalable for large datasets.