The bimodality index: a criterion for discovering and ranking bimodal signatures from cancer gene expression profiling data

A Review of Single-Cell RNA-Seq Annotation, Integration, and Cell-Cell Communication

Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful tool for investigating cellular biology at an unprecedented resolution, enabling the characterization of cellular heterogeneity, identification of rare but significant cell types, and exploration of cell-cell communications and interactions. Its broad applications span both basic and clinical research domains. In this comprehensive review, we survey the current landscape of scRNA-seq analysis methods and tools, focusing on count modeling, cell-type annotation, data integration, including spatial transcriptomics, and the inference of cell-cell communication. We review the challenges encountered in scRNA-seq analysis, including issues of sparsity or low expression, reliability of cell annotation, and assumptions in data integration, and discuss the potential impact of suboptimal clustering and differential expression analysis tools on downstream analyses, particularly in identifying cell subpopulations. Finally, we discuss recent advancements and future directions for enhancing scRNA-seq analysis. Specifically, we highlight the development of novel tools for annotating single-cell data, integrating and interpreting multimodal datasets covering transcriptomics, epigenomics, and proteomics, and inferring cellular communication networks. By elucidating the latest progress and innovation, we provide a comprehensive overview of the rapidly advancing field of scRNA-seq analysis.

Comparison of cancer subtype identification methods combined with feature selection methods in omics data analysis

BACKGROUND

Cancer subtype identification is important for the early diagnosis of cancer and the provision of adequate treatment. Prior to identifying the subtype of cancer in a patient, feature selection is also crucial for reducing the dimensionality of the data by detecting genes that contain important information about the cancer subtype. Numerous cancer subtyping methods have been developed, and their performance has been compared. However, combinations of feature selection and subtype identification methods have rarely been considered. This study aimed to identify the best combination of variable selection and subtype identification methods in single omics data analysis.

RESULTS

Combinations of six filter-based methods and six unsupervised subtype identification methods were investigated using The Cancer Genome Atlas (TCGA) datasets for four cancers. The number of features selected varied, and several evaluation metrics were used. Although no single combination was found to have a distinctively good performance, Consensus Clustering (CC) and Neighborhood-Based Multi-omics Clustering (NEMO) used with variance-based feature selection had a tendency to show lower p-values, and nonnegative matrix factorization (NMF) stably showed good performance in many cases unless the Dip test was used for feature selection. In terms of accuracy, the combination of NMF and similarity network fusion (SNF) with Monte Carlo Feature Selection (MCFS) and Minimum-Redundancy Maximum Relevance (mRMR) showed good overall performance. NMF always showed among the worst performances without feature selection in all datasets, but performed much better when used with various feature selection methods. iClusterBayes (ICB) had decent performance when used without feature selection.

CONCLUSIONS

Rather than a single method clearly emerging as optimal, the best methodology was different depending on the data used, the number of features selected, and the evaluation method. A guideline for choosing the best combination method under various situations is provided.

Generation and multi-dimensional profiling of a childhood cancer cell line atlas defines new therapeutic opportunities

Pediatric solid and central nervous system tumors are the leading cause of cancer-related death among children. Identifying new targeted therapies necessitates the use of pediatric cancer models that faithfully recapitulate the patient's disease. However, the generation and characterization of pediatric cancer models has significantly lagged behind adult cancers, underscoring the urgent need to develop pediatric-focused cell line resources. Herein, we establish a single-site collection of 261 cell lines, including 224 pediatric cell lines representing 18 distinct extracranial and brain childhood tumor types. We subjected 182 cell lines to multi-omics analyses (DNA sequencing, RNA sequencing, DNA methylation), and in parallel performed pharmacological and genetic CRISPR-Cas9 loss-of-function screens to identify pediatric-specific treatment opportunities and biomarkers. Our work provides insight into specific pathway vulnerabilities in molecularly defined pediatric tumor classes and uncovers biomarker-linked therapeutic opportunities of clinical relevance. Cell line data and resources are provided in an open access portal.

Gene expression variability in long-term survivors of childhood cancer and cancer-free controls in response to ionizing irradiation

BACKGROUND

Differential expression analysis is usually adjusted for variation. However, most studies that examined the expression variability (EV) have used computations affected by low expression levels and did not examine healthy tissue. This study aims to calculate and characterize an unbiased EV in primary fibroblasts of childhood cancer survivors and cancer-free controls (N0) in response to ionizing radiation.

METHODS

Human skin fibroblasts of 52 donors with a first primary neoplasm in childhood (N1), 52 donors with at least one second primary neoplasm (N2 +), as well as 52 N0 were obtained from the KiKme case-control study and exposed to a high (2 Gray) and a low dose (0.05 Gray) of X-rays and sham- irradiation (0 Gray). Genes were then classified as hypo-, non-, or hyper-variable per donor group and radiation treatment, and then examined for over-represented functional signatures.

RESULTS

We found 22 genes with considerable EV differences between donor groups, of which 11 genes were associated with response to ionizing radiation, stress, and DNA repair. The largest number of genes exclusive to one donor group and variability classification combination were all detected in N0: hypo-variable genes after 0 Gray (n = 49), 0.05 Gray (n = 41), and 2 Gray (n = 38), as well as hyper-variable genes after any dose (n = 43). While after 2 Gray positive regulation of cell cycle was hypo-variable in N0, (regulation of) fibroblast proliferation was over-represented in hyper-variable genes of N1 and N2+. In N2+, 30 genes were uniquely classified as hyper-variable after the low dose and were associated with the ERK1/ERK2 cascade. For N1, no exclusive gene sets with functions related to the radiation response were detected in our data.

CONCLUSION

N2+ showed high degrees of variability in pathways for the cell fate decision after genotoxic insults that may lead to the transfer and multiplication of DNA-damage via proliferation, where apoptosis and removal of the damaged genome would have been appropriate. Such a deficiency could potentially lead to a higher vulnerability towards side effects of exposure to high doses of ionizing radiation, but following low-dose applications employed in diagnostics, as well.

Genome-wide cross-cancer analysis illustrates the critical role of bimodal miRNA in patient survival and drug responses to PI3K inhibitors

Heterogeneity of cancer means many tumorigenic genes are only aberrantly expressed in a subset of patients and thus follow a bimodal distribution, having two modes of expression within a single population. Traditional statistical techniques that compare sample means between cancer patients and healthy controls fail to detect bimodally expressed genes. We utilize a mixture modeling approach to identify bimodal microRNA (miRNA) across cancers, find consistent sources of heterogeneity, and identify potential oncogenic miRNA that may be used to guide personalized therapies. Pathway analysis was conducted using target genes of the bimodal miRNA to identify potential functional implications in cancer. In vivo overexpression experiments were conducted to elucidate the clinical importance of bimodal miRNA in chemotherapy treatments. In nine types of cancer, tumors consistently displayed greater bimodality than normal tissue. Specifically, in liver and lung cancers, high expression of miR-105 and miR-767 was indicative of poor prognosis. Functional pathway analysis identified target genes of miR-105 and miR-767 enriched in the phosphoinositide-3-kinase (PI3K) pathway, and analysis of over 200 cancer drugs in vitro showed that drugs targeting the same pathway had greater efficacy in cell lines with high miR-105 and miR-767 levels. Overexpression of the two miRNA facilitated response to PI3K inhibitor treatment. We demonstrate that while cancer is marked by considerable genetic heterogeneity, there is between-cancer concordance regarding the particular miRNA that are more variable. Bimodal miRNA are ideal biomarkers that can be used to stratify patients for prognosis and drug response in certain types of cancer.

Bimodal genes can be defined as those having two modes of expression within the same population. A variety of statistical methodologies have been employed to assess bimodal gene expression, but current methods and their applications have been limited. Given the advances in next-generation sequencing as well as the extensive regulatory role of miRNA, assessing bimodality in miRNA-seq data can greatly broaden our understanding of factors underlying tumor progression. The goal of the current study was to utilize a novel mixture modeling approach to identify bimodal miRNA and then demonstrate their importance in cancer by evaluating their ability to predict overall survival and drug response. Our results showed that high levels of bimodal miRNA expression was characteristic of cancer. Additionally, several bimodal miRNA were common to multiple cancer types, suggesting that certain miRNA consistently account for tumor heterogeneity and may be involved in general oncogenic processes. Our study points to the potential of bimodal miRNA to facilitate precise prognostic evaluation and effective treatment strategies.

Bimodal gene expression in cancer patients provides interpretable biomarkers for drug sensitivity

Identifying biomarkers predictive of cancer cell response to drug treatment constitutes one of the main challenges in precision oncology. Recent large-scale cancer pharmacogenomic studies have opened new avenues of research to develop predictive biomarkers by profiling thousands of human cancer cell lines at the molecular level and screening them with hundreds of approved drugs and experimental chemical compounds. Many studies have leveraged these data to build predictive models of response using various statistical and machine learning methods. However, a common pitfall to these methods is the lack of interpretability as to how they make predictions, hindering the clinical translation of these models. To alleviate this issue, we used the recent logic modeling approach to develop a new machine learning pipeline that explores the space of bimodally expressed genes in multiple large in vitro pharmacogenomic studies and builds multivariate, nonlinear, yet interpretable logic-based models predictive of drug response. The performance of this approach was showcased in a compendium of the three largest in vitro pharmacogenomic data sets to build robust and interpretable models for 101 drugs that span 17 drug classes with high validation rates in independent datasets. These results along with in vivo and clinical validation, support a better translation of gene expression biomarkers between model systems using bimodal gene expression.

Functional Genomic Identification of Predictors of Sensitivity and Mechanisms of Resistance to Multivalent Second-Generation TRAIL-R2 Agonists

Multivalent second-generation TRAIL-R2 agonists are currently in late preclinical development and early clinical trials. Herein, we use a representative second-generation agent, MEDI3039, to address two major clinical challenges facing these agents: lack of predictive biomarkers to enable patient selection and emergence of resistance. Genome-wide CRISPR knockout screens were notable for the lack of resistance mechanisms beyond the canonical TRAIL-R2 pathway (caspase-8, FADD, BID) as well as p53 and BAX in TP53 wild-type models, whereas a CRISPR activatory screen identified cell death inhibitors MCL-1 and BCL-XL as mechanisms to suppress MEDI3039-induced cell death. High-throughput drug screening failed to identify genomic alterations associated with response to MEDI3039; however, transcriptomics analysis revealed striking association between MEDI3039 sensitivity and expression of core components of the extrinsic apoptotic pathway, most notably its main apoptotic effector caspase-8 in solid tumor cell lines. Further analyses of colorectal cell lines and patient-derived xenografts identified caspase-8 expression ratio to its endogenous regulator FLIP(L) as predictive of sensitivity to MEDI3039 in several major solid tumor types and a further subset indicated by caspase-8:MCL-1 ratio. Subsequent MEDI3039 combination screening of TRAIL-R2, caspase-8, FADD, and BID knockout models with 60 compounds with varying mechanisms of action identified two inhibitor of apoptosis proteins (IAP) that exhibited strong synergy with MEDI3039 that could reverse resistance only in BID-deleted models. In summary, we identify the ratios of caspase-8:FLIP(L) and caspase-8:MCL-1 as potential predictive biomarkers for second-generation TRAIL-R2 agonists and loss of key effectors such as FADD and caspase-8 as likely drivers of clinical resistance in solid tumors.

A Bimodal Model Based on Truncation Positive Normal with Application to Height Data

In this work, we propose a new bimodal distribution with support in the real line. We obtain some properties of the model, such as moments, quantiles, and mode, among others. The computational implementation of the model is presented in the tpn package of the software R. We perform a simulation study in order to assess the properties of the maximum likelihood estimators in finite samples. Finally, we present an application to a bimodal data set, where our proposal is compared with other models in the literature.

On Generalizing Sarle’s Bimodality Coefficient as a Path towards a Newly Composite Bimodality Coefficient

Determining whether a distribution is bimodal is of great interest for many applications. Several tests have been developed, but the only ones that can be run extremely fast, in constant time on any variable-size signal window, are based on Sarle’s bimodality coefficient. We propose in this paper a generalization of this coefficient, to prove its validity, and show how each coefficient can be computed in a fast manner, in constant time, for random regions pertaining to a large dataset. We present some of the caveats of these coefficients and potential ways to circumvent them. We also propose a composite bimodality coefficient obtained as a product of the weighted generalized coefficients. We determine the potential best set of weights to associate with our composite coefficient when using up to three generalized coefficients. Finally, we prove that the composite coefficient outperforms any individual generalized coefficient.

Label-free sensing of cells with fluorescence lifetime imaging: The quest for metabolic heterogeneity

Molecular, morphological, and physiological heterogeneity is the inherent property of cells which governs differences in their response to external influence. Tumor cell metabolic heterogeneity is of a special interest due to its clinical relevance to tumor progression and therapeutic outcomes. Rapid, sensitive, and noninvasive assessment of metabolic heterogeneity of cells is a great demand for biomedical sciences. Fluorescence lifetime imaging (FLIM), which is an all-optical technique, is an emerging tool for sensing and quantifying cellular metabolism by measuring fluorescence decay parameters of endogenous fluorophores, such as NAD(P)H. To achieve accurate discrimination between metabolically diverse cellular subpopulations, appropriate approaches to FLIM data collection and analysis are needed. In this paper, the unique capability of FLIM to attain the overarching goal of discriminating metabolic heterogeneity is demonstrated. This has been achieved using an approach to data analysis based on the nonparametric analysis, which revealed a much better sensitivity to the presence of metabolically distinct subpopulations compared to more traditional approaches of FLIM measurements and analysis. The approach was further validated for imaging cultured cancer cells treated with chemotherapy. These results pave the way for accurate detection and quantification of cellular metabolic heterogeneity using FLIM, which will be valuable for assessing therapeutic vulnerabilities and predicting clinical outcomes.

SLFN11 biomarker status predicts response to lurbinectedin as a single agent and in combination with ATR inhibition in small cell lung cancer

Background

Lurbinectedin recently received FDA accelerated approval as a second line treatment option for metastatic small cell lung cancer (SCLC). However, there are currently no established biomarkers to predict SCLC sensitivity or resistance to lurbinectedin or preclinical studies to guide rational combinations.

Methods

Drug sensitivity was assayed in proliferation assays and xenograft models. Baseline proteomic profiling was performed by reverse-phase protein array. Lurbinectedin-induced changes in intracellular signaling pathways were assayed by Western blot.

Results

Among 21 human SCLC cell lines, cytotoxicity was observed following lurbinectedin treatment at a low dose (median IC50 0.46 nM, range, 0.06–1.83 nM). Notably, cell lines with high expression of Schlafen-11 (SLFN11) protein, a promising biomarker of response to other DNA damaging agents (e.g., chemotherapy, PARP inhibitors), were more sensitive to single-agent lurbinectedin (FC =3.2, P=0.005). SLFN11 was validated as a biomarker of sensitivity to lurbinectedin using siRNA knockdown and in xenografts representing SLFN11 high and low SCLC. Replication stress and DNA damage markers (e.g., γH2AX, phosphorylated CHK1, phosphorylated RPA32) increased in SCLC cell lines following treatment with lurbinectedin. Lurbinectedin also induced PD-L1 expression via cGAS-STING pathway activation. Finally, the combination of lurbinectedin with the ataxia telangiectasia and Rad3-related protein (ATR) inhibitors ceralasertib and berzosertib showed a greater than additive effect in SLFN11-low models.

Conclusions

Together our data confirm the activity of lurbinectedin across a large cohort of SCLC models and identify SLFN11 as a top candidate biomarker for lurbinectedin sensitivity. In SLFN11-low SCLC cell lines which are relatively resistance to lurbinectedin, the addition of an ATR inhibitor to lurbinectedin re-sensitized otherwise resistant cells, confirming previous observations that SLFN11 is a master regulator of DNA damage response independent of ATR, and the absence of SLFN11 leads to synthetic lethality with ATR inhibition. This study provides a rationale for lurbinectedin in combination with ATR inhibitors to overcome resistance in SCLC with low SLFN11 expression.

The Cancer Surfaceome Atlas integrates genomic, functional and drug response data to identify actionable targets

Cell-surface proteins (SPs) are a rich source of immune and targeted therapies. By systematically integrating single-cell and bulk genomics, functional studies and target actionability, in the present study we comprehensively identify and annotate genes encoding SPs (GESPs) pan-cancer. We characterize GESP expression patterns, recurrent genomic alterations, essentiality, receptor-ligand interactions and therapeutic potential. We also find that mRNA expression of GESPs is cancer-type specific and positively correlates with protein expression, and that certain GESP subgroups function as common or specific essential genes for tumor cell growth. We also predict receptor-ligand interactions substantially deregulated in cancer and, using systems biology approaches, we identify cancer-specific GESPs with therapeutic potential. We have made this resource available through the Cancer Surfaceome Atlas ( http://fcgportal.org/TCSA ) within the Functional Cancer Genome data portal.

A benchmark study of simulation methods for single-cell RNA sequencing data

Single-cell RNA-seq (scRNA-seq) data simulation is critical for evaluating computational methods for analysing scRNA-seq data especially when ground truth is experimentally unattainable. The reliability of evaluation depends on the ability of simulation methods to capture properties of experimental data. However, while many scRNA-seq data simulation methods have been proposed, a systematic evaluation of these methods is lacking. We develop a comprehensive evaluation framework, SimBench, including a kernel density estimation measure to benchmark 12 simulation methods through 35 scRNA-seq experimental datasets. We evaluate the simulation methods on a panel of data properties, ability to maintain biological signals, scalability and applicability. Our benchmark uncovers performance differences among the methods and highlights the varying difficulties in simulating data characteristics. Furthermore, we identify several limitations including maintaining heterogeneity of distribution. These results, together with the framework and datasets made publicly available as R packages, will guide simulation methods selection and their future development.

Assessing Consistency Across Functional Screening Datasets in Cancer Cells

Motivation

Many high-throughput screening studies have been carried out in cancer cell lines to identify therapeutic agents and targets. Existing consistency assessment studies only examined two datasets at a time, with conclusions based on a subset of carefully selected features rather than considering global consistency of all the data. However, poor concordance can still be observed for a large part of the data even when selected features are highly consistent.

Results

In this study, we assembled nine compound screening datasets and three functional genomics datasets. We derived direct measures of consistency as well as indirect measures of consistency based on association between functional data and copy number-adjusted gene expression data. These results have been integrated into a web application—the Functional Data Consistency Explorer (FDCE), to allow users to make queries and generate interactive visualizations so that functional data consistency can be assessed for individual features of interest.

Availability and implementation

The FDCE web tool and we have developed and the functional data consistency measures we have generated are available at https://lccl.shinyapps.io/FDCE/.

Supplementary information

Supplementary data are available at Bioinformatics online.

Gene expression signature predicts relapse in adult patients with cytogenetically normal acute myeloid leukemia

Although ∼80% of adult patients with cytogenetically normal acute myeloid leukemia (CN-AML) achieve a complete remission (CR), more than half of them relapse. Better identification of patients who are likely to relapse can help to inform clinical decisions. We performed RNA sequencing on pretreatment samples from 268 adults with de novo CN-AML who were younger than 60 years of age and achieved a CR after induction treatment with standard "7+3" chemotherapy. After filtering for genes whose expressions were associated with gene mutations known to impact outcome (ie, CEBPA, NPM1, and FLT3-internal tandem duplication [FLT3-ITD]), we identified a 10-gene signature that was strongly predictive of patient relapse (area under the receiver operating characteristics curve [AUC], 0.81). The signature consisted of 7 coding genes (GAS6, PSD3, PLCB4, DEXI, JMY, NRP1, C10orf55) and 3 long noncoding RNAs. In multivariable analysis, the 10-gene signature was strongly associated with relapse (P < .001), after adjustment for the FLT3-ITD, CEBPA, and NPM1 mutational status. Validation of the expression signature in an independent patient set from The Cancer Genome Atlas showed the signature's strong predictive value, with AUC = 0.78. Implementation of the 10-gene signature into clinical prognostic stratification could be useful for identifying patients who are likely to relapse.

Umpire 2.0: Simulating realistic, mixed-type, clinical data for machine learning

The Umpire 2.0 R-package offers a streamlined, user-friendly workflow to simulate complex, heterogeneous, mixed-type data with known subgroup identities, dichotomous outcomes, and time-to-event data, while providing ample opportunities for fine-tuning and flexibility. Here, we describe how we have expanded the core Umpire 1.0 R-package, developed to simulate gene expression data, to generate clinically realistic, mixed-type data for use in evaluating unsupervised and supervised machine learning (ML) methods. As the availability of large-scale clinical data for ML has increased, clinical data has posed unique challenges, including widely variable size, individual biological heterogeneity, data collection and measurement noise, and mixed data types. Developing and validating ML methods for clinical data requires data sets with known ground truth, generated from simulation. Umpire 2.0 addresses challenges to simulating realistic clinical data by providing the user a series of modules to generate survival parameters and subgroups, apply meaningful additive noise, and discretize to single or mixed data types. Umpire 2.0 provides broad functionality across sample sizes, feature spaces, and data types, allowing the user to simulate correlated, heterogeneous, binary, continuous, categorical, or mixed type data from the scale of a small clinical trial to data on thousands of patients drawn from electronic health records. The user may generate elaborate simulations by varying parameters in order to compare algorithms or interrogate operating characteristics of an algorithm in both supervised and unsupervised ML.

Integrating imaging and RNA-seq improves outcome prediction in cervical cancer

Approaches using a single type of data have been applied to classify human tumors. Here we integrate imaging features and transcriptomic data using a prospectively collected tumor bank. We demonstrate that increased maximum standardized uptake value on pretreatment 18F-fluorodeoxyglucose-positron emission tomography correlates with epithelial-to-mesenchymal transition (EMT) gene expression. We derived and validated 3 major molecular groups, namely squamous epithelial, squamous mesenchymal, and adenocarcinoma, using prospectively collected institutional (n = 67) and publicly available (n = 304) data sets. Patients with tumors of the squamous mesenchymal subtype showed inferior survival outcomes compared with the other 2 molecular groups. High mesenchymal gene expression in cervical cancer cells positively correlated with the capacity to form spheroids and with resistance to radiation. CaSki organoids were radiation-resistant but sensitive to the glycolysis inhibitor, 2-DG. These experiments provide a strategy for response prediction by integrating large data sets, and highlight the potential for metabolic therapy to influence EMT phenotypes in cervical cancer.

Comparison of Methods for Feature Selection in Clustering of High-Dimensional RNA-Sequencing Data to Identify Cancer Subtypes

Cancer subtype identification is important to facilitate cancer diagnosis and select effective treatments. Clustering of cancer patients based on high-dimensional RNA-sequencing data can be used to detect novel subtypes, but only a subset of the features (e.g., genes) contains information related to the cancer subtype. Therefore, it is reasonable to assume that the clustering should be based on a set of carefully selected features rather than all features. Several feature selection methods have been proposed, but how and when to use these methods are still poorly understood. Thirteen feature selection methods were evaluated on four human cancer data sets, all with known subtypes (gold standards), which were only used for evaluation. The methods were characterized by considering mean expression and standard deviation (SD) of the selected genes, the overlap with other methods and their clustering performance, obtained comparing the clustering result with the gold standard using the adjusted Rand index (ARI). The results were compared to a supervised approach as a positive control and two negative controls in which either a random selection of genes or all genes were included. For all data sets, the best feature selection approach outperformed the negative control and for two data sets the gain was substantial with ARI increasing from (-0.01, 0.39) to (0.66, 0.72), respectively. No feature selection method completely outperformed the others but using the dip-rest statistic to select 1000 genes was overall a good choice. The commonly used approach, where genes with the highest SDs are selected, did not perform well in our study.

The shape of gene expression distributions matter: how incorporating distribution shape improves the interpretation of cancer transcriptomic data

BACKGROUND

In genomics, we often assume that continuous data, such as gene expression, follow a specific kind of distribution. However we rarely stop to question the validity of this assumption, or consider how broadly applicable it may be to all genes that are in the transcriptome. Our study investigated the prevalence of a range of gene expression distributions in three different tumor types from the Cancer Genome Atlas (TCGA).

RESULTS

Surprisingly, the expression of less than 50% of all genes was Normally-distributed, with other distributions including Gamma, Bimodal, Cauchy, and Lognormal also represented. Most of the distribution categories contained genes that were significantly enriched for unique biological processes. Different assumptions based on the shape of the expression profile were used to identify genes that could discriminate between patients with good versus poor survival. The prognostic marker genes that were identified when the shape of the distribution was accounted for reflected functional insights into cancer biology that were not observed when standard assumptions were applied. We showed that when multiple types of distributions were permitted, i.e. the shape of the expression profile was used, the statistical classifiers had greater predictive accuracy for determining the prognosis of a patient versus those that assumed only one type of gene expression distribution.

CONCLUSIONS

Our results highlight the value of studying a gene's distribution shape to model heterogeneity of transcriptomic data and the impact on using analyses that permit more than one type of gene expression distribution. These insights would have been overlooked when using standard approaches that assume all genes follow the same type of distribution in a patient cohort.

Copula modeling of receiver operating characteristic and predictiveness curves

Receiver operating characteristic (ROC) and predictiveness curves are graphical tools to study the discriminative and predictive power of a continuous-valued marker in a binary outcome. In this paper, a copula-based construction of the joint density of the marker and the outcome is developed for plotting and analyzing both curves. The methodology only requires a copula function, the marginal distribution of the marker, and the prevalence rate for the model to be characterized. The adoption of the Gaussian copula and the customization of the margin for the marker are proposed for such characterization. The computation of both curves is numerically more feasible than methods that attempt to obtain one curve in terms of the other. Estimation is carried out using maximum likelihood and resampling-based methods. Randomized quantile residuals from each conditional distribution are employed for both assessing the adequacy of the model and identifying outliers. The performance of the estimators of both curves and their underlying quantities is evaluated in simulation studies that assume different dependence structures and sample sizes. The methods are illustrated with an analysis of the level of progesterone receptor gene expression for the diagnosis and prediction of estrogen receptor-positive breast cancer.

scClassify: sample size estimation and multiscale classification of cells using single and multiple reference

Automated cell type identification is a key computational challenge in single-cell RNA-sequencing (scRNA-seq) data. To capitalise on the large collection of well-annotated scRNA-seq datasets, we developed scClassify, a multiscale classification framework based on ensemble learning and cell type hierarchies constructed from single or multiple annotated datasets as references. scClassify enables the estimation of sample size required for accurate classification of cell types in a cell type hierarchy and allows joint classification of cells when multiple references are available. We show that scClassify consistently performs better than other supervised cell type classification methods across 114 pairs of reference and testing data, representing a diverse combination of sizes, technologies and levels of complexity, and further demonstrate the unique components of scClassify through simulations and compendia of experimental datasets. Finally, we demonstrate the scalability of scClassify on large single-cell atlases and highlight a novel application of identifying subpopulations of cells from the Tabula Muris data that were unidentified in the original publication. Together, scClassify represents state-of-the-art methodology in automated cell type identification from scRNA-seq data.

Single-cell analyses reveal increased intratumoral heterogeneity after the onset of therapy resistance in small-cell lung cancer

The natural history of small cell lung cancer (SCLC) includes rapid evolution from chemosensitivity to chemoresistance, although mechanisms underlying this evolution remain obscure due to scarcity of post-relapse tissue samples. We generated circulating tumor cell (CTC)-derived xenografts (CDXs) from SCLC patients to study intratumoral heterogeneity (ITH) via single-cell RNAseq of chemo-sensitive and -resistant CDXs and patient CTCs. We found globally increased ITH including heterogeneous expression of therapeutic targets and potential resistance pathways, such as EMT, between cellular subpopulations following treatment-resistance. Similarly, serial profiling of patient CTCs directly from blood confirmed increased ITH post-relapse. These data suggest that treatment-resistance in SCLC is characterized by coexisting subpopulations of cells with heterogeneous gene expression leading to multiple, concurrent resistance mechanisms. These findings emphasize the need for clinical efforts to focus on rational combination therapies for treatment-naïve SCLC tumors to maximize initial responses and counteract the emergence of ITH and diverse resistance mechanisms.

Bimodal age distribution at diagnosis in breast cancer persists across molecular and genomic classifications

PURPOSE

Female breast cancer demonstrates bimodal age frequency distribution patterns at diagnosis, interpretable as two main etiologic subtypes or groupings of tumors with shared risk factors. While RNA-based methods including PAM50 have identified well-established clinical subtypes, age distribution patterns at diagnosis as a proxy for etiologic subtype are not established for molecular and genomic tumor classifications.

METHODS

We evaluated smoothed age frequency distributions at diagnosis for Carolina Breast Cancer Study cases within immunohistochemistry-based and RNA-based expression categories. Akaike information criterion (AIC) values compared the fit of single density versus two-component mixture models. Two-component mixture models estimated the proportion of early-onset and late-onset categories by immunohistochemistry-based ER (n = 2860), and by RNA-based ESR1 and PAM50 subtype (n = 1965). PAM50 findings were validated using pooled publicly available data (n = 8103).

RESULTS

Breast cancers were best characterized by bimodal age distribution at diagnosis with incidence peaks near 45 and 65 years, regardless of molecular characteristics. However, proportional composition of early-onset and late-onset age distributions varied by molecular and genomic characteristics. Higher ER-protein and ESR1-RNA categories showed a greater proportion of late age-at-onset. Similarly, PAM50 subtypes showed a shifting age-at-onset distribution, with most pronounced early-onset and late-onset peaks found in Basal-like and Luminal A, respectively.

CONCLUSIONS

Bimodal age distribution at diagnosis was detected in the Carolina Breast Cancer Study, similar to national cancer registry data. Our data support two fundamental age-defined etiologic breast cancer subtypes that persist across molecular and genomic characteristics. Better criteria to distinguish etiologic subtypes could improve understanding of breast cancer etiology and contribute to prevention efforts.

Combination Olaparib and Temozolomide in Relapsed Small-Cell Lung Cancer

Small-cell lung cancer (SCLC) is an aggressive malignancy in which inhibitors of PARP have modest single-agent activity. We performed a phase I/II trial of combination olaparib tablets and temozolomide (OT) in patients with previously treated SCLC. We established a recommended phase II dose of olaparib 200 mg orally twice daily with temozolomide 75 mg/m² daily, both on days 1 to 7 of a 21-day cycle, and expanded to a total of 50 patients. The confirmed overall response rate was 41.7% (20/48 evaluable); median progression-free survival was 4.2 months [95% confidence interval (CI), 2.8-5.7]; and median overall survival was 8.5 months (95% CI, 5.1-11.3). Patient-derived xenografts (PDX) from trial patients recapitulated clinical OT responses, enabling a 32-PDX coclinical trial. This revealed a correlation between low basal expression of inflammatory-response genes and cross-resistance to both OT and standard first-line chemotherapy (etoposide/platinum). These results demonstrate a promising new therapeutic strategy in SCLC and uncover a molecular signature of those tumors most likely to respond. SIGNIFICANCE: We demonstrate substantial clinical activity of combination olaparib/temozolomide in relapsed SCLC, revealing a promising new therapeutic strategy for this highly recalcitrant malignancy. Through an integrated coclinical trial in PDXs, we then identify a molecular signature predictive of response to OT, and describe the common molecular features of cross-resistant SCLC.See related commentary by Pacheco and Byers, p. 1340.This article is highlighted in the In This Issue feature, p. 1325.

A comparison of survival analysis methods for cancer gene expression RNA-Sequencing data

Identifying genetic biomarkers of patient survival remains a major goal of large-scale cancer profiling studies. Using gene expression data to predict the outcome of a patient's tumor makes biomarker discovery a compelling tool for improving patient care. As genomic technologies expand, multiple data types may serve as informative biomarkers, and bioinformatic strategies have evolved around these different applications. For categorical variables such as a gene's mutation status, biomarker identification to predict survival time is straightforward. However, for continuous variables like gene expression, the available methods generate highly-variable results, and studies on best practices are lacking. We investigated the performance of eight methods that deal specifically with continuous data. K-means, Cox regression, concordance index, D-index, 25th-75th percentile split, median-split, distribution-based splitting, and KaplanScan were applied to four RNA-sequencing (RNA-seq) datasets from the Cancer Genome Atlas. The reliability of the eight methods was assessed by splitting each dataset into two groups and comparing the overlap of the results. Gene sets that had been identified from the literature for a specific tumor type served as positive controls to assess the accuracy of each biomarker using receiver operating characteristic (ROC) curves. Artificial RNA-Seq data were generated to test the robustness of these methods under fixed levels of gene expression noise. Our results show that methods based on dichotomizing tend to have consistently poor performance while C-index, D-index, and k-means perform well in most settings. Overall, the Cox regression method had the strongest performance based on tests of accuracy, reliability, and robustness.

Personalization of Logical Models With Multi-Omics Data Allows Clinical Stratification of Patients

Logical models of cancer pathways are typically built by mining the literature for relevant experimental observations. They are usually generic as they apply for large cohorts of individuals. As a consequence, they generally do not capture the heterogeneity of patient tumors and their therapeutic responses. We present here a novel framework, referred to as PROFILE, to tailor logical models to a particular biological sample such as a patient tumor. This methodology permits to compare the model simulations to individual clinical data, i.e., survival time. Our approach focuses on integrating mutation data, copy number alterations (CNA), and expression data (transcriptomics or proteomics) to logical models. These data need first to be either binarized or set between 0 and 1, and can then be incorporated in the logical model by modifying the activity of the node, the initial conditions or the state transition rates. The use of MaBoSS, a tool based on Monte-Carlo kinetic algorithm to perform stochastic simulations on logical models results in model state probabilities, and allows for a semi-quantitative study of the model phenotypes and perturbations. As a proof of concept, we use a published generic model of cancer signaling pathways and molecular data from METABRIC breast cancer patients. For this example, we test several combinations of data incorporation and discuss that, with these data, the most comprehensive patient-specific cancer models are obtained by modifying the nodes' activity of the model with mutations, in combination or not with CNA data, and altering the transition rates with RNA expression. We conclude that these model simulations show good correlation with clinical data such as patients' Nottingham prognostic index (NPI) subgrouping and survival time. We observe that two highly relevant cancer phenotypes derived from personalized models, Proliferation and Apoptosis, are biologically consistent prognostic factors: patients with both high proliferation and low apoptosis have the worst survival rate, and conversely. Our approach aims to combine the mechanistic insights of logical modeling with multi-omics data integration to provide patient-relevant models. This work leads to the use of logical modeling for precision medicine and will eventually facilitate the choice of patient-specific drug treatments by physicians.

The rise of the distributions: why non-normality is important for understanding the transcriptome and beyond

The application of statistics has been instrumental in clarifying our understanding of the genome. While insights have been derived for almost all levels of genome function, most importantly, statistics has had the greatest impact on improving our knowledge of transcriptional regulation. But the drive to extract the most meaningful inferences from big data can often force us to overlook the fundamental role that statistics plays, and specifically, the basic assumptions that we make about big data. Normality is a statistical property that is often swept up into an assumption that we may or may not be consciously aware of making. This review highlights the inherent value of non-normal distributions to big data analysis by discussing use cases of non-normality that focus on gene expression data. Collectively, these examples help to motivate the premise of why at this stage, now more than ever, non-normality is important for learning about gene regulation, transcriptomics, and more.

Computational methods to identify bimodal gene expression and facilitate personalized treatment in cancer patients

Standard methods for detecting cancer-associated genes rely on comparison of sample means between cancer patients and healthy controls. While such methods have successfully identified several oncogenes and tumor suppressor genes, they neglect to account for heterogeneity within the cancer population. Genetic mutations, translocations, and amplifications are often inconsistent across tumors, and instead they often affect smaller subsets of patients. This concept gives rise to the idea of bimodally expressed genes, or genes that display two modes of expression within one population. Analysis of bimodal gene expression has been explored via a variety of techniques including test statistics and clustering. In this review, we summarize the methodologies used to quantify bimodal gene expression and address the utility of these genes in patient stratification and specialized therapeutics in breast and lung cancer. Finally we discuss the limitations and future directions for bimodal genes in the era of high-throughput sequencing and personalized medicine.

Single-cell RNA-Seq analysis identifies a noncoding interleukin 4 (IL-4) RNA that post-transcriptionally up-regulates IL-4 production in T helper cells

High-throughput sequencing has revealed a tremendous complexity of cellular transcriptomes, which is partly due to the generation of multiple alternative transcripts from a single gene locus. Because alternative transcripts often have low abundance in bulk cells, the functions of most of these transcripts and their relationship with their canonical counterparts remain unclear. Here we applied single-cell RNA-Seq to analyze the transcriptome complexity of in vitro-differentiated, murine type 2 T helper (Th2) cells. We found that cytokine gene transcripts contribute most of the intercellular heterogeneity, with a group of universal cytokines, including interleukins 1a, 2, 3, and 16, being bimodally expressed. At the single-cell level, use of alternative promoters prevalently generated alternative transcripts. For instance, although undetectable in bulk cells, a noncoding RNA isoform of IL-4 (IL4nc), which was driven by an intronic promoter in the IL-4 locus, was predominantly expressed in a subset of Th2 cells. IL4nc displayed distinct temporal expression patterns compared with the canonical IL-4 mRNA and post-transcriptionally promoted the production of IL-4 protein in Th2 cells. In conclusion, our findings reveal a mechanism whereby minor noncanonical transcripts post-transcriptionally regulate expression of their cognate canonical genes.

Dynamic variations in epithelial-to-mesenchymal transition (EMT), ATM, and SLFN11 govern response to PARP inhibitors and cisplatin in small cell lung cancer

Small cell lung cancer (SCLC) is one of the most aggressive forms of cancer, with a 5-year survival <7%. A major barrier to progress is the absence of predictive biomarkers for chemotherapy and novel targeted agents such as PARP inhibitors. Using a high-throughput, integrated proteomic, transcriptomic, and genomic analysis of SCLC patient-derived xenografts (PDXs) and profiled cell lines, we identified biomarkers of drug sensitivity and determined their prevalence in patient tumors. In contrast to breast and ovarian cancer, PARP inhibitor response was not associated with mutations in homologous recombination (HR) genes (e.g., BRCA1/2) or HRD scores. Instead, we found several proteomic markers that predicted PDX response, including high levels of SLFN11 and E-cadherin and low ATM. SLFN11 and E-cadherin were also significantly associated with in vitro sensitivity to cisplatin and topoisomerase1/2 inhibitors (all commonly used in SCLC). Treatment with cisplatin or PARP inhibitors downregulated SLFN11 and E-cadherin, possibly explaining the rapid development of therapeutic resistance in SCLC. Supporting their functional role, silencing SLFN11 reduced in vitro sensitivity and drug-induced DNA damage; whereas ATM knockdown or pharmacologic inhibition enhanced sensitivity. Notably, SCLC with mesenchymal phenotypes (i.e., loss of E-cadherin and high epithelial-to-mesenchymal transition (EMT) signature scores) displayed striking alterations in expression of miR200 family and key SCLC genes (e.g., NEUROD1, ASCL1, ALDH1A1, MYCL1). Thus, SLFN11, EMT, and ATM mediate therapeutic response in SCLC and warrant further clinical investigation as predictive biomarkers.

Statin-Induced Cancer Cell Death Can Be Mechanistically Uncoupled from Prenylation of RAS Family Proteins

The statin family of drugs preferentially triggers tumor cell apoptosis by depleting mevalonate pathway metabolites farnesyl pyrophosphate (FPP) and geranylgeranyl pyrophosphate (GGPP), which are used for protein prenylation, including the oncoproteins of the RAS superfamily. However, accumulating data indicate that activation of the RAS superfamily are poor biomarkers of statin sensitivity, and the mechanism of statin-induced tumor-specific apoptosis remains unclear. Here we demonstrate that cancer cell death triggered by statins can be uncoupled from prenylation of the RAS superfamily of oncoproteins. Ectopic expression of different members of the RAS superfamily did not uniformly sensitize cells to fluvastatin, indicating that increased cellular demand for protein prenylation cannot explain increased statin sensitivity. Although ectopic expression of HRAS increased statin sensitivity, expression of myristoylated HRAS did not rescue this effect. HRAS-induced epithelial-to-mesenchymal transition (EMT) through activation of zinc finger E-box binding homeobox 1 (ZEB1) sensitized tumor cells to the antiproliferative activity of statins, and induction of EMT by ZEB1 was sufficient to phenocopy the increase in fluvastatin sensitivity; knocking out ZEB1 reversed this effect. Publicly available gene expression and statin sensitivity data indicated that enrichment of EMT features was associated with increased sensitivity to statins in a large panel of cancer cell lines across multiple cancer types. These results indicate that the anticancer effect of statins is independent from prenylation of RAS family proteins and is associated with a cancer cell EMT phenotype.Significance: The use of statins to target cancer cell EMT may be useful as a therapy to block cancer progression. Cancer Res; 78(5); 1347-57. ©2017 AACR.

Targeted BMI1 inhibition impairs tumor growth in lung adenocarcinomas with low CEBPα expression

Lung cancer is the most common cause of cancer deaths. The expression of the transcription factor C/EBPα (CCAAT/enhancer binding protein α) is frequently lost in non-small cell lung cancer, but the mechanisms by which C/EBPα suppresses tumor formation are not fully understood. In addition, no pharmacological therapy is available to specifically target C/EBPα expression. We discovered a subset of pulmonary adenocarcinoma patients in whom negative/low C/EBPα expression and positive expression of the oncogenic protein BMI1 (B lymphoma Mo-MLV insertion region 1 homolog) have prognostic value. We also generated a lung-specific mouse model of C/EBPα deletion that develops lung adenocarcinomas, which are prevented by Bmi1 haploinsufficiency. BMI1 activity is required for both tumor initiation and maintenance in the C/EBPα-null background, and pharmacological inhibition of BMI1 exhibits antitumor effects in both murine and human adenocarcinoma lines. Overall, we show that C/EBPα is a tumor suppressor in lung cancer and that BMI1 is required for the oncogenic process downstream of C/EBPα loss. Therefore, anti-BMI1 pharmacological inhibition may offer a therapeutic benefit for lung cancer patients with low expression of C/EBPα and high BMI1.

Circulating cytokines and small molecules follow distinct expression patterns in acute myeloid leukaemia

Background:

Although aberrant expression of cytokines and small molecules (analytes) is well documented in acute myeloid leukaemia (AML), their co-expression patterns are not yet identified. In addition, plasma baselines for some analytes that are biomarkers for other cancers have not been previously reported in AML.

Methods:

We used multiplex array technology to simultaneously detect and quantify 32 plasma analyte (22 reported analytes and 10 novel analytes) levels in 38 patients.

Results:

In our study, 16 analytes are found to be significantly deregulated (13 higher, 3 lower, Mann–Whitney U-test, P-value <0.005), where 5 of them have never been reported before in AML. We predicted a seven-analyte-containing multiplex panel for diagnosis of AML and, among them, MIF could be a possible therapeutic target. In addition, we observed that circulating analytes show five co-expression signatures.

Conclusions:

Circulating analyte expression in AML significantly differs from normal, and follow distinct expression patterns.

Protein expression of TTF1 and cMYC define distinct molecular subgroups of small cell lung cancer with unique vulnerabilities to aurora kinase inhibition, DLL3 targeting, and other targeted therapies

Small cell lung cancer (SCLC) is a recalcitrant cancer for which no new treatments have been approved in over 30 years. While molecular subtyping now guides treatment selection for patients with non-small cell lung cancer and other cancers, SCLC is still treated as a single disease entity. Using model-based clustering, we found two major proteomic subtypes of SCLC characterized by either high thyroid transcription factor-1 (TTF1)/low cMYC protein expression or high cMYC/low TTF1. Applying "drug target constellation" (DTECT) mapping, we further show that protein levels of TTF1 and cMYC predict response to targeted therapies including aurora kinase, Bcl2, and HSP90 inhibitors. Levels of TTF1 and DLL3 were also highly correlated in preclinical models and patient tumors. TTF1 (used in the diagnosis lung cancer) could therefore be used as a surrogate of DLL3 expression to identify patients who may respond to the DLL3 antibody-drug conjugate rovalpituzumab tesirine. These findings suggest that TTF1, cMYC or other protein markers identified here could be used to identify subgroups of SCLC patients who may respond preferentially to several emerging targeted therapies.

Flipping between Polycomb repressed and active transcriptional states introduces noise in gene expression

Polycomb repressive complexes (PRCs) are important histone modifiers, which silence gene expression; yet, there exists a subset of PRC-bound genes actively transcribed by RNA polymerase II (RNAPII). It is likely that the role of Polycomb repressive complex is to dampen expression of these PRC-active genes. However, it is unclear how this flipping between chromatin states alters the kinetics of transcription. Here, we integrate histone modifications and RNAPII states derived from bulk ChIP-seq data with single-cell RNA-sequencing data. We find that Polycomb repressive complex-active genes have greater cell-to-cell variation in expression than active genes, and these results are validated by knockout experiments. We also show that PRC-active genes are clustered on chromosomes in both two and three dimensions, and interactions with active enhancers promote a stabilization of gene expression noise. These findings provide new insights into how chromatin regulation modulates stochastic gene expression and transcriptional bursting, with implications for regulation of pluripotency and development.Polycomb repressive complexes modify histones but it is unclear how changes in chromatin states alter kinetics of transcription. Here, the authors use single-cell RNAseq and ChIPseq to find that actively transcribed genes with Polycomb marks have greater cell-to-cell variation in expression.

CHK1 Inhibition in Small-Cell Lung Cancer Produces Single-Agent Activity in Biomarker-Defined Disease Subsets and Combination Activity with Cisplatin or Olaparib

Effective targeted therapies for small-cell lung cancer (SCLC), the most aggressive form of lung cancer, remain urgently needed. Here we report evidence of preclinical efficacy evoked by targeting the overexpressed cell-cycle checkpoint kinase CHK1 in SCLC. Our studies employed RNAi-mediated attenuation or pharmacologic blockade with the novel second-generation CHK1 inhibitor prexasertib (LY2606368), currently in clinical trials. In SCLC models in vitro and in vivo, LY2606368 exhibited strong single-agent efficacy, augmented the effects of cisplatin or the PARP inhibitor olaparib, and improved the response of platinum-resistant models. Proteomic analysis identified CHK1 and MYC as top predictive biomarkers of LY2606368 sensitivity, suggesting that CHK1 inhibition may be especially effective in SCLC with MYC amplification or MYC protein overexpression. Our findings provide a preclinical proof of concept supporting the initiation of a clinical efficacy trial in patients with platinum-sensitive or platinum-resistant relapsed SCLC. Cancer Res; 77(14); 3870-84. ©2017 AACR.

Computational identification of mutually exclusive transcriptional drivers dysregulating metastatic microRNAs in prostate cancer

Androgen-ablation therapies, which are the standard treatment for metastatic prostate cancer, invariably lead to acquired resistance. Hence, a systematic identification of additional drivers may provide useful insights into the development of effective therapies. Numerous microRNAs that are critical for metastasis are dysregulated in metastatic prostate cancer, but the underlying molecular mechanism is poorly understood. We perform an integrative analysis of transcription factor (TF) and microRNA expression profiles and computationally identify three master TFs, AR, HOXC6 and NKX2-2, which induce the aberrant metastatic microRNA expression in a mutually exclusive fashion. Experimental validations confirm that the three TFs co-dysregulate a large number of metastasis-associated microRNAs. Moreover, their overexpression substantially enhances cell motility and is consistently associated with a poor clinical outcome. Finally, the mutually exclusive overexpression between AR, HOXC6 and NKX2-2 is preserved across various tissues and cancers, suggesting that mutual exclusivity may represent an intrinsic characteristic of driver TFs during tumorigenesis.

Dysregulation of microRNAs is thought to be important for metastasis in castration-resistant prostate cancer (CRPC), but its drivers are unknown. Here, the authors use computational analysis to identify HOXC6 and NKX2-2 in addition to AR as alternative drivers of dysregulated miRNA expression in CRPC.

Further evidence to support bimodality of oestrogen receptor expression in breast cancer

AIMS

Although oestrogen receptor (ER)-negative breast cancers (BCs) do not respond to hormone therapy, the response of ER-positive BCs is reported to be variable, which may suggest a dose-dependent effect. The aim of this study was to assess the pattern of ER expression in BCs at the protein (immunohistochemistry) and transcriptome (microarray-based gene expression) levels.

METHODS AND RESULTS

ER immunohistochemical (IHC) expression was assessed in a large series of BCs, including 3649 core biopsies and 1892 cases prepared as tissue microarrays (TMAs) stained with specific antibodies. ESR1 mRNA expression was assessed in the METABRIC study (1980 cases), by the use of the Linear Models for Microarray Data (limma) software, and the results were compared with protein levels. IHC data confirmed the bimodality of ER expression, with 92.2% and 89.2% of the cases showing completely negative (<1%) or highly positive (≥70%) expression on the cores and TMAs, respectively. Weakly positive cases (1-10%) and intermediately positive (11-69%) cases were infrequent (2.7% and 5.1%, and 1.6% and 9.2%, in cores and TMAs, respectively), and did not show survival difference from ER-negative tumours. When full-face sections of the corresponding excision specimens were immunostained, 47% of the ER-low/intermediate group were deemed to be ER-negative. Transcriptomic data not only showed a significant correlation between ESR1 mRNA and protein expression levels, but also confirmed the bimodality of ER expression at the mRNA level.

CONCLUSIONS

Our study provides further evidence that ER expression is bimodal, and that it is observed at both the mRNA and protein levels. The reported poor survival of BC patients with low ER expression in the early clinical trials may be related to the inclusion of ER-negative cases.

A statistical approach for identifying differential distributions in single-cell RNA-seq experiments

The ability to quantify cellular heterogeneity is a major advantage of single-cell technologies. However, statistical methods often treat cellular heterogeneity as a nuisance. We present a novel method to characterize differences in expression in the presence of distinct expression states within and among biological conditions. We demonstrate that this framework can detect differential expression patterns under a wide range of settings. Compared to existing approaches, this method has higher power to detect subtle differences in gene expression distributions that are more complex than a mean shift, and can characterize those differences. The freely available R package scDD implements the approach.

Clinically Viable Gene Expression Assays with Potential for Predicting Benefit from MEK Inhibitors

Purpose: To develop a clinically viable gene expression assay to measure RAS/RAF/MEK/ERK (RAS-ERK) pathway output suitable for hypothesis testing in non-small cell lung cancer (NSCLC) clinical studies.Experimental Design: A published MEK functional activation signature (MEK signature) that measures RAS-ERK functional output was optimized for NSCLC in silico NanoString assays were developed for the NSCLC optimized MEK signature and the 147-gene RAS signature. First, platform transfer from Affymetrix to NanoString, and signature modulation following treatment with KRAS siRNA and MEK inhibitor, were investigated in cell lines. Second, the association of the signatures with KRAS mutation status, dynamic range, technical reproducibility, and spatial and temporal variation was investigated in NSCLC formalin-fixed paraffin-embedded tissue (FFPET) samples.Results: We observed a strong cross-platform correlation and modulation of signatures in vitro Technical and biological replicates showed consistent signature scores that were robust to variation in input total RNA; conservation of scores between primary and metastatic tumor was statistically significant. There were statistically significant associations between high MEK (P = 0.028) and RAS (P = 0.003) signature scores and KRAS mutation in 50 NSCLC samples. The signatures identify overlapping but distinct candidate patient populations from each other and from KRAS mutation testing.Conclusions: We developed a technically and biologically robust NanoString gene expression assay of MEK pathway output, compatible with the quantities of FFPET routinely available. The gene signatures identified a different patient population for MEK inhibitor treatment compared with KRAS mutation testing. The predictive power of the MEK signature should be studied further in clinical trials. Clin Cancer Res; 23(6); 1471-80. ©2016 AACRSee related commentary by Xue and Lito, p. 1365.

High-Level Clonal FGFR Amplification and Response to FGFR Inhibition in a Translational Clinical Trial

FGFR1 and FGFR2 are amplified in many tumor types, yet what determines response to FGFR inhibition in amplified cancers is unknown. In a translational clinical trial, we show that gastric cancers with high-level clonal FGFR2 amplification have a high response rate to the selective FGFR inhibitor AZD4547, whereas cancers with subclonal or low-level amplification did not respond. Using cell lines and patient-derived xenograft models, we show that high-level FGFR2 amplification initiates a distinct oncogene addiction phenotype, characterized by FGFR2-mediated transactivation of alternative receptor kinases, bringing PI3K/mTOR signaling under FGFR control. Signaling in low-level FGFR1-amplified cancers is more restricted to MAPK signaling, limiting sensitivity to FGFR inhibition. Finally, we show that circulating tumor DNA screening can identify high-level clonally amplified cancers. Our data provide a mechanistic understanding of the distinct pattern of oncogene addiction seen in highly amplified cancers and demonstrate the importance of clonality in predicting response to targeted therapy.

SIGNIFICANCE

Robust single-agent response to FGFR inhibition is seen only in high-level FGFR-amplified cancers, with copy-number level dictating response to FGFR inhibition in vitro, in vivo, and in the clinic. High-level amplification of FGFR2 is relatively rare in gastric and breast cancers, and we show that screening for amplification in circulating tumor DNA may present a viable strategy to screen patients. Cancer Discov; 6(8); 838-51. ©2016 AACR.This article is highlighted in the In This Issue feature, p. 803.

Outlier analysis of functional genomic profiles enriches for oncology targets and enables precision medicine

Background

Genome-scale functional genomic screens across large cell line panels provide a rich resource for discovering tumor vulnerabilities that can lead to the next generation of targeted therapies. Their data analysis typically has focused on identifying genes whose knockdown enhances response in various pre-defined genetic contexts, which are limited by biological complexities as well as the incompleteness of our knowledge. We thus introduce a complementary data mining strategy to identify genes with exceptional sensitivity in subsets, or outlier groups, of cell lines, allowing an unbiased analysis without any a priori assumption about the underlying biology of dependency.

Results

Genes with outlier features are strongly and specifically enriched with those known to be associated with cancer and relevant biological processes, despite no a priori knowledge being used to drive the analysis. Identification of exceptional responders (outliers) may not lead only to new candidates for therapeutic intervention, but also tumor indications and response biomarkers for companion precision medicine strategies. Several tumor suppressors have an outlier sensitivity pattern, supporting and generalizing the notion that tumor suppressors can play context-dependent oncogenic roles.

Conclusions

The novel application of outlier analysis described here demonstrates a systematic and data-driven analytical strategy to decipher large-scale functional genomic data for oncology target and precision medicine discoveries.

Electronic supplementary material

The online version of this article (doi:10.1186/s12864-016-2807-y) contains supplementary material, which is available to authorized users.

Selecting Reliable mRNA Expression Measurements Across Platforms Improves Downstream Analysis

With increasing use of publicly available gene expression data sets, the quality of the expression data is a critical issue for downstream analysis, gene signature development, and cross-validation of data sets. Thus, identifying reliable expression measurements by leveraging multiple mRNA expression platforms is an important analytical task. In this study, we propose a statistical framework for selecting reliable measurements between platforms by modeling the correlations of mRNA expression levels using a beta-mixture model. The model-based selection provides an effective and objective way to separate good probes from probes with low quality, thereby improving the efficiency and accuracy of the analysis. The proposed method can be used to compare two microarray technologies or microarray and RNA sequencing measurements. We tested the approach in two matched profiling data sets, using microarray gene expression measurements from the same samples profiled on both Affymetrix and Illumina platforms. We also applied the algorithm to mRNA expression data to compare Affymetrix microarray data with RNA sequencing measurements. The algorithm successfully identified probes/genes with reliable measurements. Removing the unreliable measurements resulted in significant improvements for gene signature development and functional annotations.

Distinct breast cancer characteristics between screen- and self-detected breast cancers recorded in the Japanese Breast Cancer Registry

The rate of breast cancer screening for women of all ages in Japan is increasing. However, little is known about the biological differences between screen- and self-detected tumors. We used data from the Japanese Breast Cancer Registry (JBCR), a nationwide registry of newly diagnosed breast cancer cases in Japan, to investigate patients diagnosed between January 1, 2004 and December 31, 2011. We compared the clinicopathological features of tumors and assessed yearly trends regarding the proportion of screen-detected cases during the study period. We found that 31.8 % (65,358/205,544) of cancers were detected by screening. Asymptomatic tumors detected by screening (asymptomatic) were more likely to have favorable prognostic features than those that were self-detected (ductal carcinoma in situ [DCIS]: 19.8 versus 4.1 %, node-negative: 77.0 versus 61.6 %, and estrogen receptor-positive [ER+]: 82.0 versus 72.9 %, respectively). All these findings were statistically significant (p < .001). The proportion of breast cancers detected by screening among all cases increased from 21.7 % in 2004 to 37.1 % in 2011. During the same time period, the proportion of screen-detected DCIS increased from 41.5 to 66.0 % and that of ER+ cancers increased from 23.2 to 39.7 %. This study demonstrated that low-risk tumors, including DCIS, ER+, and lower TNM stage, account for a substantial proportion of clinical screening-detected cancers. The differences in biological characteristics between screen- and self-detected cancers may account in part for the limited efficacy of breast cancer screening programs aimed at improving breast cancer mortality.

Multi-omic measurement of mutually exclusive loss-of-function enriches for candidate synthetic lethal gene pairs

Background

Identification of synthetic lethal interactions in cancer cells could offer promising new therapeutic targets. Large-scale functional genomic screening presents an opportunity to test large numbers of cancer synthetic lethal hypotheses. Methods enriching for candidate synthetic lethal targets in molecularly defined cancer cell lines can steer effective design of screening efforts. Loss of one partner of a synthetic lethal gene pair creates a dependency on the other, thus synthetic lethal gene pairs should never show simultaneous loss-of-function. We have developed a computational approach to mine large multi-omic cancer data sets and identify gene pairs with mutually exclusive loss-of-function. Since loss-of-function may not always be genetic, we look for deleterious mutations, gene deletion and/or loss of mRNA expression by bimodality defined with a novel algorithm BiSEp.

Results

Applying this toolkit to both tumour cell line and patient data, we achieve statistically significant enrichment for experimentally validated tumour suppressor genes and synthetic lethal gene pairings. Notably non-reliance on genetic loss reveals a number of known synthetic lethal relationships otherwise missed, resulting in marked improvement over genetic-only predictions. We go on to establish biological rationale surrounding a number of novel candidate synthetic lethal gene pairs with demonstrated dependencies in published cancer cell line shRNA screens.

Conclusions

This work introduces a multi-omic approach to define gene loss-of-function, and enrich for candidate synthetic lethal gene pairs in cell lines testable through functional screens. In doing so, we offer an additional resource to generate new cancer drug target and combination hypotheses. Algorithms discussed are freely available in the BiSEp CRAN package at http://cran.r-project.org/web/packages/BiSEp/index.html.

Electronic supplementary material

The online version of this article (doi:10.1186/s12864-016-2375-1) contains supplementary material, which is available to authorized users.

Systems approaches to predict the functions of glycoside hydrolases during the life cycle of Aspergillus niger using developmental mutants ∆brlA and ∆flbA

BACKGROUND

The filamentous fungus Aspergillus niger encounters carbon starvation in nature as well as during industrial fermentations. In response, regulatory networks initiate and control autolysis and sporulation. Carbohydrate-active enzymes play an important role in these processes, for example by modifying cell walls during spore cell wall biogenesis or in cell wall degradation connected to autolysis.

RESULTS

In this study, we used developmental mutants (ΔflbA and ΔbrlA) which are characterized by an aconidial phenotype when grown on a plate, but also in bioreactor-controlled submerged cultivations during carbon starvation. By comparing the transcriptomes, proteomes, enzyme activities and the fungal cell wall compositions of a wild type A. niger strain and these developmental mutants during carbon starvation, a global overview of the function of carbohydrate-active enzymes is provided. Seven genes encoding carbohydrate-active enzymes, including cfcA, were expressed during starvation in all strains; they may encode enzymes involved in cell wall recycling. Genes expressed in the wild-type during starvation, but not in the developmental mutants are likely involved in conidiogenesis. Eighteen of such genes were identified, including characterized sporulation-specific chitinases and An15g02350, member of the recently identified carbohydrate-active enzyme family AA11. Eight of the eighteen genes were also expressed, independent of FlbA or BrlA, in vegetative mycelium, indicating that they also have a role during vegetative growth. The ΔflbA strain had a reduced specific growth rate, an increased chitin content of the cell wall and specific expression of genes that are induced in response to cell wall stress, indicating that integrity of the cell wall of strain ΔflbA is reduced.

CONCLUSION

The combination of the developmental mutants ΔflbA and ΔbrlA resulted in the identification of enzymes involved in cell wall recycling and sporulation-specific cell wall modification, which contributes to understanding cell wall remodeling mechanisms during development.