ICAT

ICAT identifies and resolves distinct cellular identities across experimental conditions in single-cell RNA sequencing (scRNA-seq) perturbation studies.

Key Features:

• Unsupervised algorithm: Employs an unsupervised approach that does not require predefined cell states or marker genes for scRNA-seq data.
• Self-supervised feature weighting: Uses self-supervised feature weighting to prioritize relevant gene expression features for distinguishing cell identities.
• Control-guided clustering: Incorporates control-guided clustering to maintain population substructure and match cell states across experimental conditions.
• Robustness and flexibility: Demonstrates robustness to low signal strength, high perturbation severity, and varying cell type proportions.

Scientific Applications:

• Perturbation scRNA-seq analysis: Identifies unique cellular responses and shifts in cell-state composition in perturbation experiments using single-cell RNA sequencing.
• Detection of perturbation-unique cell states: Resolves perturbation-unique cellular responses that can be missed by traditional integration workflows.
• Empirical validation: Validated using simulated datasets, real scRNA-seq datasets, and a developmental model.

Methodology:

Unsupervised algorithm integrating self-supervised feature weighting with control-guided clustering; validated on simulated and real datasets.