MISC: missing imputation for single-cell RNA sequencing data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for studying cell heterogeneity.
• scRNA-seq data frequently contains high percentages of missing values (up to 30%) due to low capture efficiency and stochastic gene expression.
• Accurate imputation of missing values requires identifying missing data locations, quantifying missingness, and estimating the values.

Purpose of the Study:

• To develop a novel hybrid machine learning model for accurate imputation of missing values in scRNA-seq data.
• To address the challenges of identifying missing data, quantifying missingness, and recovering missing values.

Main Methods:

• Proposed a novel model named MISC (Missing Imputation for Single-cell RNA-seq).
• Employed a hybrid machine learning approach involving binary classification for missing data location, zero-inflated and false negative models for missingness quantification, and regression for value recovery.

Main Results:

• MISC demonstrated improved cell type classification on chronic myeloid leukemia (CML) and mouse brain datasets.
• Identified a trajectory branch in CML data, providing evidence for stem cell evolution.
• Clearly delineated subpopulations in mouse brain pyramidal CA1 and grouped oligodendrocyte cells distinctly.

Conclusions:

• The MISC model effectively imputes missing data in scRNA-seq datasets.
• MISC enhances cell type classification and aids in the study of cellular heterogeneity.
• MISC is a robust tool for addressing missing data challenges in scRNA-seq analysis.