optimalFlow

optimalFlow applies optimal transport and similarity-distance metrics with Wasserstein barycenters to cluster flow cytometry datasets, generate prototype cytometries, reduce inter-sample variability, and improve supervised cell-type classification and automated gating.

Key Features:

• OptimalFlowTemplates: Uses a similarity distance metric and Wasserstein barycenters to cluster flow cytometry datasets, generate prototype cytometries that reduce within-cluster variability, and enable supervised learning on more homogeneous groups.
• OptimalFlowClassification: Employs a database of gated cytometries together with OptimalFlowTemplates to assign cell types in new flow cytometry datasets, with superior performance reported on benchmark datasets.

Scientific Applications:

• Supervised Gating: Clustering and prototype cytometries reduce variability and make supervised gating more robust across flow cytometry datasets.
• Population Matching: High-accuracy cell-type classification facilitates matching cell populations across different flow cytometry experiments.

Methodology:

OptimalFlow leverages optimal transport techniques by clustering datasets using similarity distances and Wasserstein barycenters to form representative prototypes, then focuses supervised learning and automated gating on these reduced, more consistent groups.