scAnno

scAnno annotates single-cell RNA sequencing (scRNA-seq) datasets by applying a joint deconvolution strategy combined with logistic regression to assign cell-type labels and identify marker genes for biological interpretation.

Key Features:

• Deconvolution strategy-based annotation: Employs a joint deconvolution strategy combined with logistic regression to perform automated annotation at the single-cell cluster level.
• Reference profiles: Utilizes constructed reference profiles for human (30 cell types across 50 tissues) and mouse (26 cell types across 50 tissues) to support annotation.
• Marker gene identification: Identifies cell type-specific marker genes by integrating co-expression genes with seed genes to form a core marker set with high expression specificity.
• Validation and benchmarking: Validated on peripheral blood mononuclear cell datasets with marker concordance to the CellMarker database and benchmarked against SingleR, scPred, CHETAH, and scmap-cluster, reporting average annotation accuracies of 99.05% internally and 95.56% across platforms.

Scientific Applications:

• Cell-type annotation in complex tissues: Provides precise cell-type labels for scRNA-seq datasets from heterogeneous and rare cell populations in complex tissue samples.
• Developmental biology: Supports identification of cell-type–specific transcriptional programs relevant to developmental processes using scRNA-seq data.
• Immunology and disease pathology: Enables characterization of immune cell types and disease-associated cellular heterogeneity through marker gene identification and annotation.

Methodology:

Applies joint deconvolution at the single-cell cluster level combined with logistic regression, uses human and mouse reference profiles (30 human cell types × 50 tissues; 26 mouse cell types × 50 tissues), identifies marker genes by integrating co-expression genes with seed genes, and performs internal and cross-platform validation including comparison to the CellMarker database and other annotation tools.