Collaborative Structure-Preserved Missing Data Imputation for Single-Cell RNA-Seq Clustering.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for understanding cell types and disease progression.
• Technical variability in scRNA-seq generates significant missing data, hindering accurate clustering and cell type identification.
• Existing imputation methods often fail to fully leverage the underlying biological cluster structure.

Purpose of the Study:

• To introduce ColImpute, a novel collaborative approach for structure-preserved missing data imputation in scRNA-seq data.
• To enhance cell type identification by effectively addressing missing values in scRNA-seq datasets.
• To integrate imputation and clustering into a unified framework for improved biological insights.

Main Methods:

• Developed a unified optimization framework integrating a cluster structure-preserved imputation module and a subspace clustering module.
• Employed a collaborative training strategy where imputation guides clustering and vice versa.
• Evaluated the method's effectiveness on scRNA-seq datasets for imputation accuracy and cell type identification.

Main Results:

• ColImpute effectively imputes missing values in scRNA-seq data while preserving essential cluster structures.
• The collaborative approach enhances the performance of both the imputation and clustering modules.
• Experimental results demonstrate superior performance in cell type identification compared to existing methods.

Conclusions:

• ColImpute offers a robust solution for handling missing data in scRNA-seq analysis.
• The structure-preserving imputation strategy improves the accuracy of cell type discovery.
• This method advances the utility of scRNA-seq for understanding complex biological systems and diseases.