GoM DE: interpreting structure in sequence count data with differential expression analysis allowing for grades of...

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Parts-based representations like non-negative matrix factorization and topic modeling reveal structures in single-cell sequencing data.
• These methods capture patterns not easily identified by traditional clustering or dimensionality reduction techniques.
• Interpreting the biological meaning of these identified 'parts' or topics remains a significant challenge.

Purpose of the Study:

• To develop a novel method for enhancing the interpretation of parts-based representations in single-cell data.
• To address the challenge of annotating biological meaning to the components identified by methods like topic modeling.
• To introduce Grade of Membership differential expression (GoM DE) as a tool for this purpose.

Main Methods:

• Extending existing differential expression analysis methods.
• Implementing a Grade of Membership (GoM) model that allows cells to have partial membership across multiple biological groups.
• Applying the GoM DE method to analyze single-cell RNA-seq (scRNA-seq) and single-cell ATAC-seq (scATAC-seq) data.

Main Results:

• Demonstrated the utility of GoM DE in annotating topics derived from single-cell sequencing data.
• Showcased how partial membership assignment aids in understanding the biological context of identified components.
• Successfully applied GoM DE to both scRNA-seq and scATAC-seq datasets, highlighting its versatility.

Conclusions:

• Grade of Membership differential expression (GoM DE) provides a robust framework for interpreting parts-based structures in single-cell data.
• The method facilitates more nuanced biological interpretation by allowing cells to belong to multiple groups simultaneously.
• GoM DE is a valuable advancement for annotating topics and understanding cellular heterogeneity in complex single-cell datasets.