Visualizing and Interpreting Single-Cell Gene Expression Datasets with Similarity Weighted Nonnegative Embedding.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• High-throughput single-cell gene expression profiling is revolutionizing cell type and developmental trajectory discovery.
• Traditional visualization methods like t-stochastic neighbor embedding (t-SNE) excel at local patterns but often distort global data structures, hindering comprehensive biological interpretation.
• Accurate visualization is critical for understanding complex single-cell datasets.

Purpose of the Study:

• To develop a novel visualization method that preserves both local and global structures in single-cell gene expression data.
• To enhance the interpretability of single-cell datasets by integrating cell states, genes, and biological factors into a unified visualization.
• To provide a robust tool for analyzing developmental trajectories and discrete cell types.

Main Methods:

• Developed Similarity Weighted Nonnegative Embedding (SWNE), a novel dimensionality reduction technique.
• Employed nonnegative matrix factorization to decompose gene expression matrices into biologically relevant factors.
• Integrated cells, genes, and factors into a 2D visualization, smoothed using a similarity matrix.

Main Results:

• SWNE effectively maintains fidelity in visualizing both local and global structures within single-cell datasets.
• The method successfully embeds genes and factors alongside cells, aiding in the interpretation of cell state distinctions.
• Demonstrated SWNE's utility on single-cell RNA-seq data from hematopoietic progenitors and human brain cells.

Conclusions:

• SWNE offers enhanced interpretability for single-cell gene expression data compared to existing methods.
• The approach provides a more accurate representation of global data topology, crucial for understanding developmental processes and cell type heterogeneity.
• SWNE is available as an R package, facilitating its adoption in biological research.