Theoretical framework for the difference of two negative binomial distributions and its application in comparative...

Area of Science:

• Genomics
• Computational Biology
• Statistical Genetics

Background:

• High-throughput sequencing (HTS) is crucial for biological research at bulk and single-cell levels.
• Comparative analysis of HTS data often uses the difference of two negative binomial distributions (DOTNB), but theoretical results are limited.
• Existing methods for detecting differentially expressed genes (DEGs) in single-cell RNA sequencing (scRNA-seq) data have limitations.

Purpose of the Study:

• To derive theoretical results for DOTNB and examine its asymptotic properties.
• To introduce DEGage, a novel computational method for DEG detection in scRNA-seq data.
• To validate DEGage's performance against existing DEG analysis tools.

Main Methods:

• Derivation of basic analytical results and examination of asymptotic properties for DOTNB.
• Development of DEGage, a computational tool utilizing DOTNB for DEG identification in scRNA-seq data.
• Extensive validation using simulated and real scRNA-seq datasets, comparing DEGage with DEGseq2, DEsingle, edgeR, Monocle3, and scDD.

Main Results:

• DEGage demonstrates superior performance compared to five popular DEG analysis tools.
• The method is robust against high dropout rates and shows enhanced sensitivity for both balanced and imbalanced datasets, even with small sample sizes.
• DEGage successfully identified marker genes in prostate cancer and potential memory-related genes in mouse neurons.

Conclusions:

• DEGage offers a powerful and reliable approach for DEG analysis in scRNA-seq data.
• The theoretical advancements in DOTNB and the DEGage software have broad applicability for HTS data analysis.
• This work facilitates comparative analyses of dispersed count data and addresses significant research questions in genomics and beyond.