Deterministic column subset selection for single-cell RNA-Seq.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-Seq) data analysis requires gene filtering and dimensionality reduction.
• Existing methods like PCA and thresholding have limitations, including loss of data structure and ignored correlations.

Purpose of the Study:

• To introduce and evaluate Deterministic Column Subset Selection (DCSS) for scRNA-Seq data analysis.
• To demonstrate DCSS's ability to preserve non-negativity and sparsity while addressing collinearity and covariance.

Main Methods:

• Developed and derived new spectral bounds for DCSS.
• Applied DCSS to gene expression data from two scRNA-Seq experiments.
• Compared DCSS performance against three common gene filtering thresholding methods.

Main Results:

• DCSS successfully selected a small subset of genes that yielded clustering results comparable to using the full gene set.
• Clusters derived from DCSS-selected genes were informative for cell type identification.
• DCSS demonstrated favorable properties, avoiding pitfalls of PCA and thresholding methods.

Conclusions:

• DCSS is an effective method for gene selection in scRNA-Seq data analysis.
• The method preserves crucial data structures and improves interpretability.
• DCSS facilitates accurate cell type classification from reduced gene subsets.