scACCorDiON

scACCorDiON leverages optimal transport to identify disease-related communication patterns by analyzing cell-cell communication graphs in patient cohorts.

Key Features:

• Optimal transport-based graph comparison: Explores node distances on the Markov Chain as the ground metric between directed weighted graphs representing cell-cell communication events.
• k-barycenter clustering: Derives a k-barycenter algorithm using Wasserstein-based distance to cluster directed weighted graphs.
• Explainable cluster representation: Barycenters provide a robust and explainable representation of cell-cell communication events related to detected clusters.
• Disease status prediction: Predicts clusters that match the disease status of samples in large cohorts of scRNA-seq data.

Scientific Applications:

• Cohort-based cell-cell communication analysis: Analyzes sample-specific cell-cell communication events in single-cell data produced in large patient cohorts.
• Disease sub-cluster detection: Detects sub-clusters of disease samples associated with changes in the tumor microenvironment, as demonstrated in pancreas adenocarcinoma.
• Characterization of complex tissues: Characterizes cell communication events in complex tissues using single-cell sequencing combined with ligand-receptor analysis.

Methodology:

scACCorDiON uses an optimal transport algorithm exploring node distances on the Markov Chain as the ground metric between directed weighted graphs, and derives a k-barycenter algorithm using Wasserstein-based distance for clustering.