SIMBA

SIMBA embeds single cells together with genes, chromatin-accessible regions, and DNA sequences into a shared latent space using graph embedding to enable clustering-free marker discovery, gene-regulatory inference, batch-effect removal, and integration of multi-omic single-cell datasets.

Key Features:

• Joint embedding of cells and features: SIMBA jointly embeds single cells and their defining features (genes, chromatin-accessible regions, DNA sequences) into a common latent space to represent interactions between feature types.
• Clustering-free marker discovery: Enables marker identification without relying on predefined cell clusters, supporting detection of markers across continuous cellular states.
• Gene regulation inference: Uses the integrated cell-feature representation to infer gene-regulatory relationships and putative regulatory mechanisms.
• Batch effect removal: Integrates data across batches to mitigate batch effects and improve comparability across datasets.
• Omics data integration: Supports integration of genetic, epigenetic, and other molecular single-cell omics to enable holistic analyses.
• Unified framework for diverse problems: Provides a single representational framework that can be applied across multiple single-cell modalities and analysis tasks.

Scientific Applications:

• Cellular heterogeneity analysis: Relates cell states to genes, chromatin-accessible regions, and DNA sequences in a shared embedding to dissect heterogeneity.
• Exploratory marker discovery: Identifies novel cell-type or state markers without prior clustering assumptions.
• Functional genomics and regulatory inference: Infers gene regulatory interactions and supports functional interpretation of single-cell signals.
• Multi-study and multi-omic integration: Combines datasets and modalities while addressing batch effects for integrated single-cell analyses.

Methodology:

Graph embedding techniques that jointly represent cells and their associated features in a latent space, capturing intrinsic data structure and relationships between feature types.