FRMC: a fast and robust method for the imputation of scRNA-seq data

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell transcriptome sequencing provides high-resolution gene expression data, crucial for biomedical research.
• High missing values (dropout events) in scRNA-Seq data due to technical limitations hinder accurate downstream analysis.
• Existing imputation methods have limitations, including data distribution assumptions, inability to differentiate technical vs. biological zeros, and poor computational performance.

Purpose of the Study:

• To develop a rapid, accurate, and stable imputation method for single-cell RNA sequencing (scRNA-Seq) data.
• To address the challenge of high missing values in gene-cell expression matrices.
• To facilitate accurate downstream biological mechanism analysis.

Main Methods:

• Developed FRMC, a novel imputation software for scRNA-Seq data.
• Employed a fast and accurate singular value thresholding approximation method.
• Implemented FRMC in Python for accessibility.

Main Results:

• FRMC precisely distinguishes 'true zeros' from dropout events.
• The software accurately imputes missing values caused by technical noise.
• FRMC effectively enhances intracellular and intergenic connections, leading to accurate cell clustering in biological applications.

Conclusions:

• FRMC is a powerful tool for single-cell data analysis, ensuring biological significance, accuracy, and rapidity.
• The software overcomes limitations of existing imputation methods.
• FRMC is freely accessible for non-commercial use.