Sincast

Sincast predicts cell identities in single-cell RNA-sequencing (scRNA-seq) datasets by projecting single-cell transcriptomes onto bulk reference atlases to leverage their phenotype annotations.

Key Features:

• Projection onto bulk reference atlases: Projects single-cell transcriptomes onto established bulk reference atlases to utilize their extensive phenotype annotations.
• Pseudo-bulk aggregation: Combines expression profiles of similar cells to create bulk-like profiles that mirror bulk RNA-seq characteristics.
• Graph-based imputation: Imputes missing values in sparse scRNA-seq expression profiles using a graph-based approach to enhance data completeness.
• Batch effect handling via transformation: Circumvents explicit batch effect correction by aligning single-cell data with bulk references through direct transformation methods.
• Continuous cell identity prediction: Predicts cell identities along a continuum to enable identification of novel cell states not present in reference atlases.
• Mapping into bulk-defined expression space: Positions single cells within the expression space defined by bulk references to place cells in biological niches.

Scientific Applications:

• Novel cell state discovery: Identifies cell states absent from existing reference atlases by leveraging continuum predictions.
• Annotation of scRNA-seq datasets: Annotates single-cell profiles using phenotype annotations from bulk reference atlases.
• Biological niche mapping: Projects single cells into bulk-defined expression space to locate cells within relevant biological niches.
• Downstream interpretation of transcriptomic landscapes: Facilitates analysis of cellular heterogeneity and discovery of new cell types or states using bulk-derived phenotypic information.

Methodology:

Projects single-cell transcriptomes onto bulk reference atlases using pseudo-bulk aggregation and graph-based imputation, and aligns single-cell data to bulk references via direct transformation methods to avoid explicit batch correction.