scInt

scInt integrates heterogeneous single-cell RNA sequencing (scRNA-seq) data to construct robust cell-cell similarity metrics, learn contrastive biological variation across datasets, and transfer knowledge from integrated reference to query datasets for comparative and developmental analyses.

Key Features:

• Robust Cell-Cell Similarity Construction: Constructs accurate and robust cell-cell similarity metrics to integrate datasets with confounding biological and technical variations.
• Unified Contrastive Biological Variation Learning: Leverages unified contrastive learning to capture and align biological variation across multiple scRNA-seq datasets.
• Knowledge Transfer from Reference to Query Data Sets: Facilitates transfer of learned information from an integrated reference dataset to query datasets to support downstream comparative analyses.

Scientific Applications:

• Integration Across Developmental Stages: Integrates developmental trajectories across stages, exemplified by applications to mouse developing tracheal epithelial cells.
• Identification of Condition-Specific Cell Subpopulations: Identifies functionally distinct cell subpopulations specific to biological conditions to dissect cellular heterogeneity and condition-specific responses.

Methodology:

Constructs robust cell-cell similarity metrics between cells from different scRNA-seq datasets; employs contrastive learning techniques to unify biological variation across integrated data; facilitates knowledge transfer from an integrated reference dataset to query datasets.