Imputation method for single-cell RNA-seq data using neural topic model.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) offers high resolution for studying cellular heterogeneity.
• The dropout phenomenon, characterized by excess zero values, complicates downstream functional analysis of scRNA-seq data.
• Accurate imputation of missing expression data is crucial for robust scRNA-seq analysis.

Purpose of the Study:

• To introduce scNTImpute, a novel imputation framework designed to overcome the challenges posed by dropout events in scRNA-seq data.
• To leverage a neural topic model for accurate identification and imputation of missing gene expression values.
• To enhance the biological insights derived from scRNA-seq data by improving cell subset clustering.

Main Methods:

• Developed scNTImpute, an imputation framework utilizing a neural topic model.
• Employed a neural network encoder to extract topic features and infer cell similarity.
• Integrated a mixture model learning approach to identify dropout-affected transcriptome data.
• Implemented a borrowing strategy using gene information from similar cells to impute missing values.

Main Results:

• scNTImpute accurately and efficiently identifies dropout values in scRNA-seq data.
• The framework successfully imputes missing gene expression values, preserving biological information.
• Cell subset clustering is significantly improved, mitigating the impact of technical noise.
• Demonstrated robust performance on real scRNA-seq datasets.

Conclusions:

• scNTImpute provides an effective solution for the dropout problem in scRNA-seq analysis.
• The method enhances the accuracy of cell clustering and recovers biological signals obscured by technical noise.
• scNTImpute facilitates more reliable downstream functional analysis of single-cell transcriptomic data.