Representation Methods of Transcriptomics with Applications in Neuroimmune Biology.

Area of Science:

• Computational biology
• Neuroscience
• Genomics

Background:

• Gene expression data defines cellular identity and molecular programs, but these are distinct.
• Microglia exhibit significant heterogeneity, yet transcriptomic data analysis often assumes distinct identities without clear transcriptional states.
• Current methods may not fully capture the complexity of microglia function.

Purpose of the Study:

• To compare two single-cell analysis methods: differential expression analysis (for identities) and co-expression network analysis (for molecular programs).
• To explore alternative transcriptomic representations for understanding microglia.
• To determine the most effective method for analyzing microglia function.

Main Methods:

• Applied differential expression analysis to identify cellular identities.
• Utilized co-expression network analysis to identify molecular programs.
• Compared the results of both methods for microglia transcriptomic data.

Main Results:

• Co-expression network analysis identified significant functional ontologies missed by differential expression analysis.
• Co-expression modules were consistent across different transcriptomic datasets.
• These modules suggest reducible functional programs that are context-dependent.

Conclusions:

• Co-expression network analysis is a best practice for single-cell analysis of individual cell types like microglia.
• Modeling microglia function as concurrent molecular programs is a more parsimonious and accurate approach.
• This highlights the importance of analyzing molecular programs over distinct identities for complex cell types.