CONGAS

CONGAS infers tumor subclonal populations by integrating bulk DNA-derived copy number alterations with single-cell RNA sequencing to cluster cells and identify clone-specific RNA expression phenotypes.

Key Features:

• Bayesian probabilistic framework: Constructs priors using bulk DNA sequencing data to inform clustering of single cells and does not require a normal reference.
• COpy Number Alteration-based Gaussian Approximation: Models the relationship between copy number alterations and RNA expression using a Gaussian approximation.
• Integration of independent assays: Phases bulk DNA and single-cell RNA measurements obtained from independent assays to jointly analyze CNAs and expression.
• Joint CNA–expression clustering: Clusters single cells according to CNA profiles and associated RNA expression differences to identify subclonal populations.
• Scalability and efficient inference: Employs variational inference and GPU acceleration for scalable analysis of large single-cell datasets.

Scientific Applications:

• Subclonal Composition Analysis: Determines the subclonal composition of tumors by integrating single-cell RNA sequencing with copy number alterations.
• Clone-Specific Phenotype Identification: Identifies clone-specific RNA phenotypes to link genetic CNAs with transcriptional differences within tumor subpopulations.

Methodology:

Uses a Bayesian framework that constructs priors from bulk DNA sequencing, applies a COpy Number Alteration-based Gaussian approximation, and performs joint clustering of cells by CNA profiles and RNA expression differences using variational inference with GPU acceleration; the approach does not require a normal reference.