scRNMF: An imputation method for single-cell RNA-seq data by robust and non-negative matrix factorization.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) provides high-resolution gene expression data.
• scRNA-seq data frequently exhibit dropouts, leading to missing gene expression values.
• Missing data in scRNA-seq can introduce biases and impede downstream analyses.

Purpose of the Study:

• To develop an effective imputation method for scRNA-seq data.
• To address the challenge of missing data and technical dropouts in scRNA-seq.
• To improve the accuracy and robustness of scRNA-seq data analysis.

Main Methods:

• Proposed a novel imputation method named scRNMF (single-cell RNA-seq robust non-negative matrix factorization).
• scRNMF integrates L2 loss and C-loss functions for matrix factorization.
• The C-loss function is specifically utilized to handle zero values and improve robustness against outliers.

Main Results:

• scRNMF demonstrated superior performance in imputing scRNA-seq data compared to existing state-of-the-art methods.
• The method showed stability and power across various datasets with different sizes and zero rates.
• Robust factorization was achieved by effectively handling outliers and zero values.

Conclusions:

• scRNMF is a powerful and stable tool for imputing missing data in scRNA-seq datasets.
• The proposed method enhances the reliability of gene expression analysis from scRNA-seq data.
• scRNMF offers a robust solution to overcome technical limitations in scRNA-seq data.