An NMF-based approach to discover overlooked differentially expressed gene regions from single-cell RNA-seq data

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) allows transcriptome quantification and cell type inference.
• Annotation-based analyses may miss differential expression in unannotated transcripts, such as those from aberrant splicing.

Purpose of the Study:

• To develop a novel computational approach for discovering overlooked differentially expressed (DE) gene regions.
• To complement existing annotation-based methods in scRNA-seq data analysis.

Main Methods:

• Developed an algorithm utilizing non-negative matrix factorization to decompose count data matrices.
• Quantified DE levels from the decomposed matrix and compared them to annotation-based results.
• Applied the method to human neural stem cells, mouse ES/primitive endoderm cells, and human preimplantation embryo datasets.

Main Results:

• Successfully identified several novel DE transcripts not detected by standard annotation-based methods.
• Discovered a transcript associated with neural stem/progenitor cell differentiation modulation.
• Validated the approach across diverse scRNA-seq datasets.

Conclusions:

• The developed algorithm effectively uncovers previously unannotated DE transcripts in scRNA-seq data.
• This method provides deeper insights into cell state regulation and transcript diversity.
• Enhances the comprehensive analysis of scRNA-seq data for biological discovery.