Detection of PatIent-Level distances from single cell genomics and pathomics data with Optimal Transport (PILOT)

Abstract

AbstractAlthough clinical applications represent the next challenge in single-cell genomics and digital pathology, we are still lacking computational methods to analyse single-cell and pathomics data at a patient level for finding patient trajectories associated with diseases. This is challenging as a single-cell/pathomics data is represented by clusters of cells/structures, which cannot be easily compared with other samples. We here propose PatIent Level analysis with Optimal Transport (PILOT). PILOT uses optimal transport to compute the Wasserstein distance between two single single-cell experiments. This allows us to perform unsupervised analysis at the sample level and to uncover trajectories associated with disease progression. Moreover, PILOT provides a statistical approach to delineate non-linear changes in cell populations, gene expression and tissues structures related to the disease trajectories. We evaluate PILOT and competing approaches in disease single-cell genomics and pathomics studies with up to 1.000 patients/donors and millions of cells or structures. Results demonstrate that PILOT detects disease-associated samples, cells, structures and genes from large and complex single-cell and pathomics data.