cnnImpute: missing value recovery for single cell RNA sequencing data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) reveals cellular heterogeneity but suffers from high missingness.
• These missing values complicate downstream analyses, hindering disease research.

Purpose of the Study:

• To introduce cnnImpute, a novel convolutional neural network (CNN) method for imputing missing data in scRNA-seq.
• To evaluate the accuracy and effectiveness of cnnImpute in recovering expression values and preserving cell populations.

Main Methods:

• Developed a CNN-based imputation method (cnnImpute).
• The method estimates missing probabilities and recovers expression values using a CNN model.
• Evaluated performance through comprehensive benchmarking experiments.

Main Results:

• cnnImpute accurately imputes missing values in scRNA-seq data.
• The method effectively preserves the integrity of cell clusters.
• Demonstrated superior performance compared to existing methods in benchmarking.

Conclusions:

• cnnImpute provides an accurate and scalable solution for missing data in scRNA-seq.
• This method is a valuable resource for advancing scRNA-seq data analysis and disease research.