Analysis of multi-condition single-cell data with latent embedding multivariate regression.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Analyzing gene expression differences in heterogeneous tissues is crucial for understanding biological processes.
• Current single-cell RNA sequencing (RNA-seq) analysis often relies on discrete cell type clustering, which may not fully capture biological complexity.

Purpose of the Study:

• To introduce a new computational model, latent embedding multivariate regression (LEMUR), for analyzing multi-condition single-cell RNA sequencing data.
• To provide a method that bypasses or delays discrete cell categorization, better reflecting continuous biological variation.

Main Methods:

• LEMUR integrates data from multiple conditions.
• It predicts gene expression changes based on experimental conditions and a cell's position in a latent space.
• The model identifies groups of cells with consistent differential gene expression for each gene.

Main Results:

• LEMUR was applied to diverse biological datasets, including cancer, zebrafish development, and Alzheimer's disease.
• The model demonstrated broad applicability across these different research areas.
• It successfully identified differential gene expression patterns without relying on pre-defined cell clusters.

Conclusions:

• LEMUR offers a powerful alternative to traditional clustering methods in single-cell RNA-seq analysis.
• The model's ability to handle continuous variation and integrate multi-condition data enhances biological insights.
• LEMUR shows significant potential for advancing research in various fields, from developmental biology to disease studies.