Joint representation and visualization of derailed cell states with Decipher

Area of Science:

• Computational biology
• Genomics
• Single-cell analysis

Background:

• Comparing biological conditions (e.g., disease vs. health) is crucial for insights.
• Effective computational tools for integrating single-cell genomics data across conditions are lacking.
• Characterizing transitions from normal to deviant cell states requires advanced methods.

Purpose of the Study:

• To present Decipher, a deep generative model for characterizing derailed cell-state trajectories.
• To enable joint modeling and visualization of gene expression and cell state from single-cell RNA-seq data.
• To reveal shared and disrupted cellular dynamics across normal and perturbed conditions.

Main Methods:

• Developed Decipher, a deep generative model.
• Applied Decipher to integrate and visualize single-cell RNA-seq data from normal and perturbed conditions.
• Utilized joint modeling of gene expression and cell state.

Main Results:

• Decipher effectively characterizes derailed cell-state trajectories.
• The model reveals shared and disrupted cellular dynamics across diverse biological contexts.
• Demonstrated superior performance in analyzing pancreatitis, acute myeloid leukemia, and gastric cancer data.

Conclusions:

• Decipher provides a powerful computational tool for single-cell genomics analysis across conditions.
• The model facilitates the understanding of cell-state transitions in disease.
• Decipher enhances biological discovery by integrating multi-condition single-cell data.