scMagnifier: Resolving fine-grained cell subtypes via GRN-informed perturbations and consensus clustering

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Resolving fine-grained cell subtypes in single-cell RNA sequencing (scRNA-seq) data is difficult due to technical noise and data sparsity.
• Subtle transcriptional differences between cell subtypes are often obscured, hindering accurate identification.

Purpose of the Study:

• To present scMagnifier, a novel consensus clustering framework designed to improve the resolution of cell subtypes in scRNA-seq data.
• To amplify subtle transcriptional differences using gene regulatory network (GRN)-informed in silico perturbations.

Main Methods:

• scMagnifier perturbs candidate transcription factors (TFs) and simulates post-perturbation expression profiles.
• It propagates perturbation effects through cluster-specific GRNs and integrates clustering results across multiple perturbations.
• Introduces regulatory perturbation consensus UMAP (rpcUMAP) for enhanced visualization and optimal cluster number selection.

Main Results:

• scMagnifier consistently improves the resolution and accuracy of fine-grained cell type identification in both single-batch and multi-batch datasets.
• The rpcUMAP visualization provides clearer separation between cell subtypes.
• Demonstrates compatibility with spatial transcriptomics workflows, effectively revealing tumor cell subtypes and their spatial organization in ovarian cancer.

Conclusions:

• scMagnifier is an effective computational framework for uncovering latent cell subpopulations and enhancing cell type identification in scRNA-seq data.
• The GRN-informed perturbation approach successfully amplifies subtle transcriptional differences.
• scMagnifier advances scRNA-seq analysis and spatial transcriptomics applications, particularly in cancer research.