Supervised capacity preserving mapping: a clustering guided visualization method for scRNA-seq data.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables large-scale transcriptome analysis at the cellular level.
• Existing visualization methods like t-SNE and UMAP struggle to accurately represent the geometric relationships and variances of distinct cell populations.
• Unsupervised methods often neglect clustering information, leading to misinterpretations of distances between functional states.

Purpose of the Study:

• To introduce supCPM, a novel supervised visualization method for scRNA-seq data.
• To enhance the accurate depiction of population segregation and functional transitions.
• To improve the preservation of global geometric structure and cluster variance compared to existing methods.

Main Methods:

• Developed supCPM, a supervised computational method for scRNA-seq data visualization.
• Evaluated supCPM's performance against six other visualization techniques using synthetic and real-world datasets.
• Focused on preserving global structure, separating clusters, and tracking cluster variance.

Main Results:

• supCPM demonstrated superior performance in preserving the global geometric structure of scRNA-seq data.
• The method accurately tracked cluster variance, overcoming limitations of t-SNE and UMAP.
• supCPM provided an enhanced visualization pipeline for interpreting functional cell transitions and population segregation.

Conclusions:

• supCPM offers a robust supervised approach for scRNA-seq data visualization.
• The method improves upon existing techniques by accurately representing global structure and cluster variance.
• supCPM facilitates more reliable interpretation of cellular heterogeneity and functional states.