Quantitative framework for cancer sample comparisons via single-cell trajectory alignment

Single cell technologies are widely used to characterize tumors and allow capture at high-resolution biological processes such as tumor evolution and development through trajectory assembly. Yet, frameworks for comparing trajectories at high-resolution have been lacking, particularly when in cancer where the large phenotypic heterogeneity existing between tumors prevents a downstream analysis that preserves the individual-level resolution. Here we present tuMap, a single-cell trajectory alignment framework which exploits high-dimensional expression data to align cancer single-cell samples and order them along one common developmental axis, thus allowing for their systematic comparison. We applied this framework to single-cell mass cytometry data of acute myeloid leukemia (AML) patients to quantify changes in cellular abundances at the time of cancer diagnosis, following treatment and in relapse. The robust alignment capabilities of tuMap enable us to integrate additional external datasets in AML, such as signaling alterations and single cell gene expression to build a unified understanding of disease development. tuMap enables a meaningful quantitative analysis of tumors and discovery of subtle differences that would be missed otherwise, opening the way for a high-resolution metric of an individual’s tumor development.