scAB

scAB integrates single-cell and clinically annotated bulk sequencing data to detect multiresolution, clinically relevant cell states and derive prognostic signatures.

Key Features:

• Integration of Data Types: Integrates single-cell RNA sequencing (scRNA-seq) and single-cell ATAC-seq with clinically annotated bulk sequencing data for joint analysis.
• Knowledge- and Graph-Guided Matrix Factorization: Implements a knowledge- and graph-guided matrix factorization model to enable simultaneous analysis of single-cell and bulk datasets.
• Multiresolution Cell States: Detects multiresolution (coarse- and fine-grain) phenotype-associated cell states, revealing signals not visible in single-cell data alone.
• Identification of Clinically-Relevant Cell Subsets: Identifies cancer and stromal cell subsets in live cancer single-cell RNA-seq data that exhibit stronger associations with poor survival.
• Association with Cancer Hallmarks: Links fine-grain cell subsets to distinct cancer hallmarks to elucidate underlying biological mechanisms.
• Prognostic Signatures and Survival Predictions: Generates prognostic signatures and survival predictions that outperform existing models across various cancer types.

Scientific Applications:

• Biomarker Identification: Identifies biomarkers predictive of immunotherapy response, drug efficacy, and patient survival, demonstrated in melanoma scRNA-seq and glioma single-cell ATAC-seq datasets.
• Prognosis Signatures and Survival Predictions: Produces prognosis signatures and survival predictions across multiple cancer types with reported superior performance to existing models.
• Prioritization of Clinically-Relevant Cell Subsets: Prioritizes clinically relevant cell subsets and predictive signatures using large publicly available databases to support prognosis assessment and treatment stratification.

Methodology:

Integrates single-cell genomics (scRNA-seq and scATAC-seq) with clinically annotated bulk sequencing data using a knowledge- and graph-guided matrix factorization model for simultaneous analysis of both datasets.