MLG: multilayer graph clustering for multi-condition scRNA-seq data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) offers high-resolution insights into cellular heterogeneity.
• Dimensionality reduction is crucial for integrating and analyzing multi-condition scRNA-seq data.
• Existing methods face challenges in accurately identifying novel cell types or transient states.

Purpose of the Study:

• To introduce Multilayer Graph Clustering (MLG) as an advanced integrative approach for scRNA-seq data.
• To enhance the signal-to-noise ratio in multi-condition scRNA-seq datasets.
• To improve the accuracy of cell population identification and data integration.

Main Methods:

• Developed MLG, an integrative method combining multiple dimensionality reduction techniques.
• Constructed a multilayer shared nearest neighbor cell graph.
• Benchmarked MLG against current best practices using large-scale datasets.

Main Results:

• MLG demonstrated superior clustering accuracy compared to existing methods.
• The approach effectively boosts the signal-to-noise ratio for fine-grained sub-population identification.
• MLG successfully identified novel cell types and developmental stages across diverse datasets.

Conclusions:

• MLG provides a robust and accurate method for integrating multi-condition scRNA-seq data.
• The technique enhances the discovery of cellular heterogeneity and transient states.
• MLG is broadly applicable to single-cell data integration challenges involving dimension reduction.