Self-assembling manifolds in single-cell RNA sequencing data.

Area of Science:

• Computational Biology
• Genomics
• Parasitology

Background:

• Single-cell RNA sequencing (scRNA-seq) enables cell type classification and analysis of biological processes.
• Accurate identification of biologically relevant genes is crucial but challenging for subtle cellular differences.
• Existing computational methods struggle with noisy or low-dimensional scRNA-seq datasets.

Purpose of the Study:

• To introduce the self-assembling manifold (SAM) algorithm for gene relevance quantification and dimensionality reduction in scRNA-seq data.
• To address the challenge of identifying subtle biological variations in complex single-cell datasets.
• To provide a robust computational tool for enhancing scRNA-seq data analysis.

Main Methods:

• Developed the self-assembling manifold (SAM) algorithm, an iterative soft feature selection strategy.
• Applied SAM to quantify gene relevance and improve dimensionality reduction.
• Validated SAM's performance against state-of-the-art methods using experimental and benchmark datasets.

Main Results:

• Experimental validation in *Schistosoma mansoni* identified novel stem cell populations.
• SAM demonstrated superior performance in identifying biologically relevant genes across 56 diverse scRNA-seq datasets.
• The algorithm consistently outperformed existing methods in biological and quantitative benchmarks.

Conclusions:

• The self-assembling manifold (SAM) algorithm is a generalizable and effective tool for scRNA-seq data analysis.
• SAM enhances the ability to detect subtle biological differences and improve cell type classification.
• This method has significant implications for understanding cell heterogeneity and developmental processes in various biological systems.