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

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.

An integrated approach to uncover drivers of cancer.

Systematic characterization of cancer genomes has revealed a staggering number of diverse aberrations that differ among individuals, such that the functional importance and physiological impact of most tumor genetic alterations remain poorly defined. We developed a computational framework that integrates chromosomal copy number and gene expression data for detecting aberrations that promote cancer progression. We demonstrate the utility of this framework using a melanoma data set. Our analysis correctly identified known drivers of melanoma and predicted multiple tumor dependencies. Two dependencies, TBC1D16 and RAB27A, confirmed empirically, suggest that abnormal regulation of protein trafficking contributes to proliferation in melanoma. Together, these results demonstrate the ability of integrative Bayesian approaches to identify candidate drivers with biological, and possibly therapeutic, importance in cancer.

Interferon α/ß Enhances the Cytotoxic Response of MEK Inhibition in Melanoma.

Drugs that inhibit the MAPK pathway have therapeutic benefit in melanoma, but responses vary between patients, for reasons that are still largely unknown. Here we aim at explaining this variability using pre- and post-MEK inhibition transcriptional profiles in a panel of melanoma cell lines. We found that most targets are context specific, under the influence of the pathway in only a subset of cell lines. We developed a computational method to identify context-specific targets, and found differences in the activity levels of the interferon pathway, driven by a deletion of the interferon locus. We also discovered that IFNα/ß treatment strongly enhances the cytotoxic effect of MEK inhibition, but only in cell lines with low activity of interferon pathway. Taken together, our results suggest that the interferon pathway plays an important role in, and predicts, the response to MAPK inhibition in melanoma. Our analysis demonstrates the value of system-wide perturbation data in predicting drug response.

Inference of modules associated to eQTLs.

Cataloging the association of transcripts to genetic variants in recent years holds the promise for functional dissection of regulatory structure of human transcription. Here, we present a novel approach, which aims at elucidating the joint relationships between transcripts and single-nucleotide polymorphisms (SNPs). This entails detection and analysis of modules of transcripts, each weakly associated to a single genetic variant, together exposing a high-confidence association signal between the module and this 'main' SNP. To explore how transcripts in a module are related to causative loci for that module, we represent such dependencies by a graphical model. We applied our method to the existing data on genetics of gene expression in the liver. The modules are significantly more, larger and denser than found in permuted data. Quantification of the confidence in a module as a likelihood score, allows us to detect transcripts that do not reach genome-wide significance level. Topological analysis of each module identifies novel insights regarding the flow of causality between the main SNP and transcripts. We observe similar annotations of modules from two sources of information: the enrichment of a module in gene subsets and locus annotation of the genetic variants. This and further phenotypic analysis provide a validation for our methodology.

Modularity and interactions in the genetics of gene expression.

Understanding the effect of genetic sequence variation on phenotype is a major challenge that lies at the heart of genetics. We developed GOLPH (GenOmic Linkage to PHenotype), a statistical method to identify genetic interactions, and used it to characterize the landscape of genetic interactions between gene expression quantitative trait loci. Our results reveal that allele-specific interactions, in which a gene only exerts an influence on the phenotype in the presence of a particular allele at the primary locus, are widespread and that genetic interactions are predominantly nonadditive. The data portray a complex picture in which interacting loci influence the expression of modules of coexpressed genes involved in coherent biological processes and pathways. We show that genetic variation at a single gene can have a major impact on the global transcriptional response, altering interactions between genes through shutdown or activation of pathways. Thus, different cellular states occur not only in response to the external environment but also result from intrinsic genetic variation.

Context Sensitive Modeling of Cancer Drug Sensitivity.

Recent screening of drug sensitivity in large panels of cancer cell lines provides a valuable resource towards developing algorithms that predict drug response. Since more samples provide increased statistical power, most approaches to prediction of drug sensitivity pool multiple cancer types together without distinction. However, pan-cancer results can be misleading due to the confounding effects of tissues or cancer subtypes. On the other hand, independent analysis for each cancer-type is hampered by small sample size. To balance this trade-off, we present CHER (Contextual Heterogeneity Enabled Regression), an algorithm that builds predictive models for drug sensitivity by selecting predictive genomic features and deciding which ones should-and should not-be shared across different cancers, tissues and drugs. CHER provides significantly more accurate models of drug sensitivity than comparable elastic-net-based models. Moreover, CHER provides better insight into the underlying biological processes by finding a sparse set of shared and type-specific genomic features.

Systems biology. Conditional density-based analysis of T cell signaling in single-cell data.

Cellular circuits sense the environment, process signals, and compute decisions using networks of interacting proteins. To model such a system, the abundance of each activated protein species can be described as a stochastic function of the abundance of other proteins. High-dimensional single-cell technologies, such as mass cytometry, offer an opportunity to characterize signaling circuit-wide. However, the challenge of developing and applying computational approaches to interpret such complex data remains. Here, we developed computational methods, based on established statistical concepts, to characterize signaling network relationships by quantifying the strengths of network edges and deriving signaling response functions. In comparing signaling between naïve and antigen-exposed CD4(+) T lymphocytes, we find that although these two cell subtypes had similarly wired networks, naïve cells transmitted more information along a key signaling cascade than did antigen-exposed cells. We validated our characterization on mice lacking the extracellular-regulated mitogen-activated protein kinase (MAPK) ERK2, which showed stronger influence of pERK on pS6 (phosphorylated-ribosomal protein S6), in naïve cells as compared with antigen-exposed cells, as predicted. We demonstrate that by using cell-to-cell variation inherent in single-cell data, we can derive response functions underlying molecular circuits and drive the understanding of how cells process signals.

High-resolution analysis of DNA regulatory elements by synthetic saturation mutagenesis.

We present a method that harnesses massively parallel DNA synthesis and sequencing for the high-throughput functional analysis of regulatory sequences at single-nucleotide resolution. As a proof of concept, we quantitatively assayed the effects of all possible single-nucleotide mutations for three bacteriophage promoters and three mammalian core promoters in a single experiment per promoter. The method may also serve as a rapid screening tool for regulatory element engineering in synthetic biology.