Unbiased visualization of single-cell genomic data with SCUBI

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Scatterplot visualizations are standard for analyzing single-cell genomic data.
• Existing scatterplot methods introduce biases by masking cells and being skewed by sample size.
• These biases hinder accurate interpretation of gene expression and cell-type composition.

Purpose of the Study:

• To develop an unbiased visualization method for single-cell genomic data.
• To address biases in representing gene expression/cell identity and comparing cell-type compositions across samples.
• To improve the faithfulness of low-dimensional data representation.

Main Methods:

• Developed SCUBI (Single-Cell Unbiased BI-visualization) method.
• SCUBI aggregates cell information within non-overlapping squares to mitigate masking bias.
• SCUBI visualizes differences in cell proportions across samples to address compositional bias.

Main Results:

• SCUBI provides a more faithful visual representation of single-cell RNA sequencing (RNA-seq) data.
• Demonstrated effectiveness on a real-world RNA-seq dataset.
• The method successfully addresses both identified visualization biases.

Conclusions:

• SCUBI offers a significant improvement over traditional scatterplot visualizations for single-cell genomics.
• The method has the potential to alter standard practices in visualizing low-dimensional single-cell data.
• SCUBI enhances the reliability of insights derived from single-cell analyses.