Data-Driven Phenotypic Dissection of AML Reveals Progenitor-like Cells that Correlate with Prognosis

Abstract

SUMMARY Acute myeloid leukemia (AML) manifests as phenotypically and functionally diverse cells, often within the same patient. Intratumor phenotypic and functional heterogeneity have been linked primarily by physical sorting experiments, which assume that functionally distinct subpopulations can be prospectively isolated by surface phenotypes. This assumption has proven problematic and we therefore developed a data-driven approach. Using mass cytometry, we profiled surface and intracellular signaling proteins simultaneously in millions of healthy and leukemic cells. We developed PhenoGraph, which algorithmically defines phenotypes in high-dimensional single-cell data. PhenoGraph revealed that the surface phenotypes of leukemic blasts do not necessarily reflect their intracellular state. Using hematopoietic progenitors, we defined a signaling-based measure of cellular phenotype, which led to isolation of a gene expression signature that was predictive of survival in independent cohorts. This study presents new methods for large-scale analysis of single-cell heterogeneity and demonstrates their utility, yielding insights into AML pathophysiology.

A single-cell survey of the small intestinal epithelium

“…To cluster single cells by their expression, we used an unsupervised clustering approach, based on the Infomap graph-clustering algorithm 9 , following approaches recently described for single-cell CyTOF data 57 and scRNA-seq 10 . Briefly, we constructed a k -nearest-neighbor ( k NN) graph on the data using, for each pair of cells, the Euclidean distance between the scores of significant PCs to identify k nearest neighbors.…”

The end of gating? An introduction to automated analysis of high dimensional cytometry data

“…Cluster frequencies can thus directly be compared, without manual gating on the SPADE tree or t-SNE map as performed previously, thus further automating and debiasing analysis. This makes PhenoGraph particularly well suited for the identification of potentially novel cellular clusters in unknown samples, as has been shown for AML patient samples [45].…”

“…Also, the density-peak detection algorithms mentioned above (ACCENSE and DensVM) [34,35] do not take the entire dimensionality of a given dataset into account. To do so, a novel algorithm was recently developed and termed "PhenoGraph" [45].…”

“…The resulting clusters can then be visualized on a t-SNE map, which often corresponds well with the PhenoGraph clustering, or alternatively on a heat map, which will show the expression levels of selected marker across all found clusters (Fig. 2D).The reported application of PhenoGraph to high-dimensional data of human bone marrow derived from healthy donors as well as acute myeloid leukemia (AML) patients [45] suggests that this method is superior to previously available clustering methods, and can resolve subpopulations as rare as 1 in 2000 cells. In fact, one advantage distinguishing PhenoGraph from previous dimensionality reduction approaches is the addition of a second iteration process using the so-called Jaccard similarity coefficient.…”

“…hematopoietic stem cells) that could potentially be obscured either by noise or by other larger cell populations. PhenoGraph has also been made available in the CYT plugin for MATLAB [45]. …”