SeqBMC: Single-cell data processing using iterative block matrix completion algorithm based on matrix factorisation.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• High-throughput sequencing generates large gene expression datasets.
• Single-cell RNA sequencing (scRNA-seq) analysis is crucial for understanding cellular heterogeneity.
• Missing data in scRNA-seq matrices poses challenges for downstream analysis.

Purpose of the Study:

• To develop and evaluate a novel algorithm for imputing missing values in scRNA-seq data.
• To improve the accuracy of cell type classification using completed gene expression matrices.
• To offer a machine learning-based solution requiring minimal biological prior knowledge.

Main Methods:

• Proposed an iterative block matrix completion algorithm (SeqBMC) based on matrix factorization.
• Utilized gene frequency to block the matrix for efficient processing.
• Applied matrix factorization to complete smaller blocks while preserving biological zeros.

Main Results:

• SeqBMC significantly enhanced gene expression matrix classification performance, achieving an 86.81% F1 score.
• The algorithm demonstrated superior performance compared to ALRA, with a 5.47% increase in accuracy and a 5.03% increase in F1 score.
• The method effectively aids in cell type recognition from scRNA-seq data.

Conclusions:

• SeqBMC offers a robust and effective machine learning approach for matrix completion in scRNA-seq data.
• The algorithm's ability to handle missing data improves the reliability of cell type identification.
• SeqBMC presents significant advantages over existing methods for scRNA-seq data imputation.