Automatic Classification of Cellular Expression by Nonlinear Stochastic Embedding (ACCENSE)

Abstract

Significance Mass cytometry enables the measurement of nearly 40 different proteins at the single-cell level, providing an unprecedented level of multidimensional information. Because of the complexity of these datasets across diverse populations of cells, new computational tools are needed to glean useful biological insights. Here we describe ACCENSE (Automatic Classification of Cellular Expression by Nonlinear Stochastic Embedding), a tool that computes a two-dimensional nonlinear distillation of the raw data, and automatically stratifies cells into phenotypic subpopulations based on their distribution of markers. Applying this tool to murine CD8 + T-cell data recovers known naive and memory subpopulations, and reveals additional diversity within these. In particular, we identify a novel subpopulation with a distinct multivariate phenotype, but which is not distinguishable on a biaxial plot of conventional markers.

Keywords

Affiliated Institutions

• Stanford University US
• Ragon Institute of MGH, MIT and Harvard US
• Massachusetts Institute of Technology US

Related Publications