Search

Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.

Abstract High throughput RNA sequencing experiments are widely conducted and analyzed to identify differentially expressed genes (DEGs). The statistical models calculated for this task are often not clear to practitioners, and analyses may not be optimally tailored to the research hypothesis. Often, interaction effects (IEs) are the mathematical equivalent of the biological research question but are not considered for different reasons. We fill this gap by explaining and presenting the potential benefit of IEs in the search for DEGs using RNA-Seq data of mice that receive different diets for different time periods. Using an IE model leads to a smaller, but likely more biologically informative set of DEGs compared to a common approach that avoids the calculation of IEs.

Abstract Background An important problem in toxicology in the context of gene expression data is the simultaneous inference of a large number of concentration–response relationships. The quality of the inference substantially depends on the choice of design of the experiments, in particular, on the set of different concentrations, at which observations are taken for the different genes under consideration. As this set has to be the same for all genes, the efficient planning of such experiments is very challenging. We address this problem by determining efficient designs for the simultaneous inference of a large number of concentration–response models. For that purpose, we both construct a D-optimality criterion for simultaneous inference and a K-means procedure which clusters the support points of the locally D-optimal designs of the individual models. Results We show that a planning of experiments that addresses the simultaneous inference of a large number of concentration–response relationships yields a substantially more accurate statistical analysis. In particular, we compare the performance of the constructed designs to the ones of other commonly used designs in terms of D-efficiencies and in terms of the quality of the resulting model fits using a real data example dealing with valproic acid. For the quality comparison we perform an extensive simulation study. Conclusions The design maximizing the D-optimality criterion for simultaneous inference improves the inference of the different concentration–response relationships substantially. The design based on the K-means procedure also performs well, whereas a log-equidistant design, which was also included in the analysis, performs poorly in terms of the quality of the simultaneous inference. Based on our findings, the D-optimal design for simultaneous inference should be used for upcoming analyses dealing with high-dimensional gene expression data.

Summary Animal studies for embryotoxicity evaluation of potential therapeutics and environmental factors are complex, costly, and time-consuming. Often, studies are not of human relevance because of species differences. In the present study, we recapitulated the process of cardiomyogenesis in human induced pluripotent stem cells (hiPSCs) by modulation of the Wnt signaling pathway to identify a key cardiomyogenesis gene signature that can be applied to identify compounds and/or stress factors compromising the cardiomyogenesis process. Among the 23 tested teratogens and 16 non-teratogens, we identified three retinoids including 13-cis-retinoic acid that completely block the process of cardiomyogenesis in hiPSCs. Moreover, we have identified an early gene signature consisting of 31 genes and associated biological processes that are severely affected by the retinoids. To predict the inhibitory potential of teratogens and non-teratogens in the process of cardiomyogenesis we established the “Developmental Cardiotoxicity Index” (CDI31g) that accurately differentiates teratogens and non-teratogens to do or do not affect the differentiation of hiPSCs to functional cardiomyocytes.

Simulation is a crucial tool for the evaluation and comparison of statistical methods. How to design fair and neutral simulation studies is therefore of great interest for both researchers developing new methods and practitioners confronted with the choice of the most suitable method. The term simulation usually refers to parametric simulation, that is, computer experiments using artificial data made up of pseudo-random numbers. Plasmode simulation, that is, computer experiments using the combination of resampling feature data from a real-life dataset and generating the target variable with a known user-selected outcome-generating model, is an alternative that is often claimed to produce more realistic data. We compare parametric and Plasmode simulation for the example of estimating the mean squared error (MSE) of the least squares estimator (LSE) in linear regression. If the true underlying data-generating process (DGP) and the outcome-generating model (OGM) were known, parametric simulation would obviously be the best choice in terms of estimating the MSE well. However, in reality, both are usually unknown, so researchers have to make assumptions: in Plasmode simulation studies for the OGM, in parametric simulation for both DGP and OGM. Most likely, these assumptions do not exactly reflect the truth. Here, we aim to find out how assumptions deviating from the true DGP and the true OGM affect the performance of parametric and Plasmode simulations in the context of MSE estimation for the LSE and in which situations which simulation type is preferable. Our results suggest that the preferable simulation method depends on many factors, including the number of features, and on how and to what extent the assumptions of a parametric simulation differ from the true DGP. Also, the resampling strategy used for Plasmode influences the results. In particular, subsampling with a small sampling proportion can be recommended.

Abstract Background In high-dimensional data (HDD) settings, the number of variables associated with each observation is very large. Prominent examples of HDD in biomedical research include omics data with a large number of variables such as many measurements across the genome, proteome, or metabolome, as well as electronic health records data that have large numbers of variables recorded for each patient. The statistical analysis of such data requires knowledge and experience, sometimes of complex methods adapted to the respective research questions. Methods Advances in statistical methodology and machine learning methods offer new opportunities for innovative analyses of HDD, but at the same time require a deeper understanding of some fundamental statistical concepts. Topic group TG9 “High-dimensional data” of the STRATOS (STRengthening Analytical Thinking for Observational Studies) initiative provides guidance for the analysis of observational studies, addressing particular statistical challenges and opportunities for the analysis of studies involving HDD. In this overview, we discuss key aspects of HDD analysis to provide a gentle introduction for non-statisticians and for classically trained statisticians with little experience specific to HDD. Results The paper is organized with respect to subtopics that are most relevant for the analysis of HDD, in particular initial data analysis, exploratory data analysis, multiple testing, and prediction. For each subtopic, main analytical goals in HDD settings are outlined. For each of these goals, basic explanations for some commonly used analysis methods are provided. Situations are identified where traditional statistical methods cannot, or should not, be used in the HDD setting, or where adequate analytic tools are still lacking. Many key references are provided. Conclusions This review aims to provide a solid statistical foundation for researchers, including statisticians and non-statisticians, who are new to research with HDD or simply want to better evaluate and understand the results of HDD analyses.

Abstract The experience of adversity in childhood has been associated with poor health outcomes in adulthood. In search of the biological mechanisms underlying these effects, research so far focused on alterations of DNA methylation or shifts in transcriptomic profiles. The level of protein, however, has been largely neglected. We utilized mass spectrometry to investigate the proteome of CD14+ monocytes in healthy adults reporting childhood adversity and a control group before and after psychosocial stress exposure. Particular proteins involved in (i) immune processes, such as neutrophil-related proteins, (ii) protein metabolism, or (iii) proteins related to mitochondrial biology, such as those involved in energy production processes, were upregulated in participants reporting exposure to adversity in childhood. This functional triad was further corroborated by protein interaction- and co-expression analyses, was independent of stress exposure, i.e. observed at both pre- and post-stress time points, and became evident especially in females. In line with the mitochondrial allostatic load model, our findings provide evidence for the long-term effects of childhood adversity on mitochondrial biology.

Abstract Background Intrinsic or acquired resistance to HER2-targeted therapy is often a problem when small molecule tyrosine kinase inhibitors or antibodies are used to treat patients with HER2 positive breast cancer. Therefore, the identification of new targets and therapies for this patient group is warranted. Activated choline metabolism, characterized by elevated levels of choline-containing compounds, has been previously reported in breast cancer. The glycerophosphodiesterase EDI3 (GPCPD1), which hydrolyses glycerophosphocholine to choline and glycerol-3-phosphate, directly influences choline and phospholipid metabolism, and has been linked to cancer-relevant phenotypes in vitro. While the importance of choline metabolism has been addressed in breast cancer, the role of EDI3 in this cancer type has not been explored. Methods EDI3 mRNA and protein expression in human breast cancer tissue were investigated using publicly-available Affymetrix gene expression microarray datasets (n = 540) and with immunohistochemistry on a tissue microarray (n = 265), respectively. A panel of breast cancer cell lines of different molecular subtypes were used to investigate expression and activity of EDI3 in vitro. To determine whether EDI3 expression is regulated by HER2 signalling, the effect of pharmacological inhibition and siRNA silencing of HER2, as well as the influence of inhibiting key components of signalling cascades downstream of HER2 were studied. Finally, the influence of silencing and pharmacologically inhibiting EDI3 on viability was investigated in vitro and on tumour growth in vivo. Results In the present study, we show that EDI3 expression is highest in ER-HER2 + human breast tumours, and both expression and activity were also highest in ER-HER2 + breast cancer cell lines. Silencing HER2 using siRNA, as well as inhibiting HER2 signalling with lapatinib decreased EDI3 expression. Pathways downstream of PI3K/Akt/mTOR and GSK3β, and transcription factors, including HIF1α, CREB and STAT3 were identified as relevant in regulating EDI3 expression. Silencing EDI3 preferentially decreased cell viability in the ER-HER2 + cells. Furthermore, silencing or pharmacologically inhibiting EDI3 using dipyridamole in ER-HER2 + cells resistant to HER2-targeted therapy decreased cell viability in vitro and tumour growth in vivo. Conclusions Our results indicate that EDI3 may be a potential novel therapeutic target in patients with HER2-targeted therapy-resistant ER-HER2 + breast cancer that should be further explored.

In toxicological concentration-response studies, a frequent goal is the determination of an 'alert concentration', i.e. the lowest concentration where a notable change in the response in comparison to the control is observed. In high-throughput gene expression experiments, e.g. based on microarray or RNA-seq technology, concentration-response profiles can be measured for thousands of genes simultaneously. One approach for determining the alert concentration is given by fitting a parametric model to the data which allows interpolation between the tested concentrations. It is well known that the quality of a model fit improves with the number of measured data points. However, adding new replicates for existing concentrations or even several replicates for new concentrations is time-consuming and expensive. Here, we propose an empirical Bayes approach to information sharing across genes, where in essence a weighted mean of the individual estimate for one specific parameter of a fitted model and the mean of all estimates of the entire set of genes is calculated as a result. Results of a controlled plasmode simulation study show that for many genes a notable improvement in terms of the mean squared error (MSE) between estimate and true underlying value of the parameter can be observed. However, for some genes, the MSE increases, and this cannot be prevented by using a more sophisticated prior distribution in the Bayesian approach.

Abstract Background Important objectives in cancer research are the prediction of a patient’s risk based on molecular measurements such as gene expression data and the identification of new prognostic biomarkers (e.g. genes). In clinical practice, this is often challenging because patient cohorts are typically small and can be heterogeneous. In classical subgroup analysis, a separate prediction model is fitted using only the data of one specific cohort. However, this can lead to a loss of power when the sample size is small. Simple pooling of all cohorts, on the other hand, can lead to biased results, especially when the cohorts are heterogeneous. Results We propose a new Bayesian approach suitable for continuous molecular measurements and survival outcome that identifies the important predictors and provides a separate risk prediction model for each cohort. It allows sharing information between cohorts to increase power by assuming a graph linking predictors within and across different cohorts. The graph helps to identify pathways of functionally related genes and genes that are simultaneously prognostic in different cohorts. Conclusions Results demonstrate that our proposed approach is superior to the standard approaches in terms of prediction performance and increased power in variable selection when the sample size is small.

Abstract Background An important task in clinical medicine is the construction of risk prediction models for specific subgroups of patients based on high-dimensional molecular measurements such as gene expression data. Major objectives in modeling high-dimensional data are good prediction performance and feature selection to find a subset of predictors that are truly associated with a clinical outcome such as a time-to-event endpoint. In clinical practice, this task is challenging since patient cohorts are typically small and can be heterogeneous with regard to their relationship between predictors and outcome. When data of several subgroups of patients with the same or similar disease are available, it is tempting to combine them to increase sample size, such as in multicenter studies. However, heterogeneity between subgroups can lead to biased results and subgroup-specific effects may remain undetected. Methods For this situation, we propose a penalized Cox regression model with a weighted version of the Cox partial likelihood that includes patients of all subgroups but assigns them individual weights based on their subgroup affiliation. The weights are estimated from the data such that patients who are likely to belong to the subgroup of interest obtain higher weights in the subgroup-specific model. Results Our proposed approach is evaluated through simulations and application to real lung cancer cohorts, and compared to existing approaches. Simulation results demonstrate that our proposed model is superior to standard approaches in terms of prediction performance and variable selection accuracy when the sample size is small. Conclusions The results suggest that sharing information between subgroups by incorporating appropriate weights into the likelihood can increase power to identify the prognostic covariates and improve risk prediction.

Wir stellen in diesem Aufsatz ein Modell interdisziplinärer Zusammenarbeit zwischen Kommunikationswissenschaft und Methodenwissenschaft (hier: Statistik) vor. Dabei steht die Frage im Mittelpunkt, wie sich die Kollaboration grundverschiedener Disziplinen über einen längeren Zeitraum verstetigen lässt. Der agilen Entwicklung von Forschungssoftware, die die fachübergreifende Zusammenarbeit strukturiert, kann hierbei eine zentrale Rolle zukommen. Sie ermöglicht zwei Ebenen der Zusammenarbeit mit verschiedenen Zeithorizonten: auf der einen werden zeitlich begrenzte Projekte umgesetzt, auf der anderen die längerfristige Zusammenarbeit organisiert. Wechselnde Team-Konstellationen und Hierarchien ermöglichen kontinuierlichen Wissensaustausch, die verwendete Programmiersprache ist Grundlage für die Entstehung einer fächerübergreifenden Kommunikation und die Orientierung auf ein gemeinsames Produkt schafft eine Kultur der gemeinsamen Verantwortlichkeit. Um die theoretischen Reflexionen an der Forschungspraxis zu spiegeln, greifen die Autor*innen auf ihre Erfahrungen im Kontext des Dortmund Center für datenbasierte Medien-Analyse (kurz: DoCMA) zurück. DoCMA ist deswegen als Fallbeispiel relevant, da Struktur und Organisation hier auf die Erfordernisse der agilen Entwicklung offener Forschungssoftware ausgerichtet wurden.

Abstract Background With modern methods in biotechnology, the search for biomarkers has advanced to a challenging statistical task exploring high dimensional data sets. Feature selection is a widely researched preprocessing step to handle huge numbers of biomarker candidates and has special importance for the analysis of biomedical data. Such data sets often include many input features not related to the diagnostic or therapeutic target variable. A less researched, but also relevant aspect for medical applications are costs of different biomarker candidates. These costs are often financial costs, but can also refer to other aspects, for example the decision between a painful biopsy marker and a simple urine test. In this paper, we propose extensions to two feature selection methods to control the total amount of such costs: greedy forward selection and genetic algorithms. In comprehensive simulation studies of binary classification tasks, we compare the predictive performance, the run-time and the detection rate of relevant features for the new proposed methods and five baseline alternatives to handle budget constraints. Results In simulations with a predefined budget constraint, our proposed methods outperform the baseline alternatives, with just minor differences between them. Only in the scenario without an actual budget constraint, our adapted greedy forward selection approach showed a clear drop in performance compared to the other methods. However, introducing a hyperparameter to adapt the benefit-cost trade-off in this method could overcome this weakness. Conclusions In feature cost scenarios, where a total budget has to be met, common feature selection algorithms are often not suitable to identify well performing subsets for a modelling task. Adaptations of these algorithms such as the ones proposed in this paper can help to tackle this problem.

In bottom-up proteomics, proteins are enzymatically digested into peptides before measurement with mass spectrometry. The relationship between proteins and their corresponding peptides can be represented by bipartite graphs. We conduct a comprehensive analysis of bipartite graphs using quantified peptides from measured data sets as well as theoretical peptides from an in silico digestion of the corresponding complete taxonomic protein sequence databases. The aim of this study is to characterize and structure the different types of graphs that occur and to compare them between data sets. We observed a large influence of the accepted minimum peptide length during in silico digestion. When changing from theoretical peptides to measured ones, the graph structures are subject to two opposite effects. On the one hand, the graphs based on measured peptides are on average smaller and less complex compared to graphs using theoretical peptides. On the other hand, the proportion of protein nodes without unique peptides, which are a complicated case for protein inference and quantification, is considerably larger for measured data. Additionally, the proportion of graphs containing at least one protein node without unique peptides rises when going from database to quantitative level. The fraction of shared peptides and proteins without unique peptides as well as the complexity and size of the graphs highly depends on the data set and organism. Large differences between the structures of bipartite peptide-protein graphs have been observed between database and quantitative level as well as between analyzed species. In the analyzed measured data sets, the proportion of protein nodes without unique peptides ranged from 6.4% to 55.0%. This highlights the need for novel methods that can quantify proteins without unique peptides. The knowledge about the structure of the bipartite peptide-protein graphs gained in this study will be useful for the development of such algorithms.

Proteasome inhibition is associated with parkinsonian pathology in vivo and degeneration of dopaminergic neurons in vitro. We explored here the metabolome (386 metabolites) and transcriptome (3257 transcripts) regulations of human LUHMES neurons, following exposure to MG-132 [100 nM]. This proteasome inhibitor killed cells within 24 h but did not reduce viability for 12 h. Overall, 206 metabolites were changed in live neurons. The early (3 h) metabolome changes suggested a compromised energy metabolism. For instance, AMP, NADH and lactate were up-regulated, while glycolytic and citric acid cycle intermediates were down-regulated. At later time points, glutathione-related metabolites were up-regulated, most likely by an early oxidative stress response and activation of NRF2/ATF4 target genes. The transcriptome pattern confirmed proteostatic stress (fast up-regulation of proteasome subunits) and also suggested the progressive activation of additional stress response pathways. The early ones (e.g., HIF-1, NF-kB, HSF-1) can be considered a cytoprotective cellular counter-regulation, which maintained cell viability. For instance, a very strong up-regulation of AIFM2 (=FSP1) may have prevented fast ferroptotic death. For most of the initial period, a definite life–death decision was not taken, as neurons could be rescued for at least 10 h after the start of proteasome inhibition. Late responses involved p53 activation and catabolic processes such as a loss of pyrimidine synthesis intermediates. We interpret this as a phase of co-occurrence of protective and maladaptive cellular changes. Altogether, this combined metabolomics–transcriptomics analysis informs on responses triggered in neurons by proteasome dysfunction that may be targeted by novel therapeutic intervention in Parkinson’s disease.

Human-relevant tests to predict developmental toxicity are urgently needed. A currently intensively studied approach makes use of differentiating human stem cells to measure chemically-induced deviations of the normal developmental program, as in a recent study based on cardiac differentiation (UKK2). Here, we (i) tested the performance of an assay modeling neuroepithelial differentiation (UKN1), and (ii) explored the benefit of combining assays (UKN1 and UKK2) that model different germ layers. Substance-induced cytotoxicity and genome-wide expression profiles of 23 teratogens and 16 non-teratogens at human-relevant concentrations were generated and used for statistical classification, resulting in accuracies of the UKN1 assay of 87–90%. A comparison to the UKK2 assay (accuracies of 90–92%) showed, in general, a high congruence in compound classification that may be explained by the fact that there was a high overlap of signaling pathways. Finally, the combination of both assays improved the prediction compared to each test alone, and reached accuracies of 92–95%. Although some compounds were misclassified by the individual tests, we conclude that UKN1 and UKK2 can be used for a reliable detection of teratogens in vitro, and that a combined analysis of tests that differentiate hiPSCs into different germ layers and cell types can even further improve the prediction of developmental toxicants.

In health research, statistical methods are frequently used to address a wide variety of research questions. For almost every analytical challenge, different methods are available. But how do we choose between different methods and how do we judge whether the chosen method is appropriate for our specific study? Like in any science, in statistics, experiments can be run to find out which methods should be used under which circumstances. The main objective of this paper is to demonstrate that simulation studies, that is, experiments investigating synthetic data with known properties, are an invaluable tool for addressing these questions. We aim to provide a first introduction to simulation studies for data analysts or, more generally, for researchers involved at different levels in the analyses of health data, who (1) may rely on simulation studies published in statistical literature to choose their statistical methods and who, thus, need to understand the criteria of assessing the validity and relevance of simulation results and their interpretation; and/or (2) need to understand the basic principles of designing statistical simulations in order to efficiently collaborate with more experienced colleagues or start learning to conduct their own simulations. We illustrate the implementation of a simulation study and the interpretation of its results through a simple example inspired by recent literature, which is completely reproducible using the R-script available from online supplemental file 1.

Mouse models of non-alcoholic fatty liver disease (NAFLD) are required to define therapeutic targets, but detailed time-resolved studies to establish a sequence of events are lacking. Here, we fed male C57Bl/6N mice a Western or standard diet over 48 weeks. Multiscale time-resolved characterization was performed using RNA-seq, histopathology, immunohistochemistry, intravital imaging, and blood chemistry; the results were compared to human disease. Acetaminophen toxicity and ammonia metabolism were additionally analyzed as functional readouts. We identified a sequence of eight key events: formation of lipid droplets; inflammatory foci; lipogranulomas; zonal reorganization; cell death and replacement proliferation; ductular reaction; fibrogenesis; and hepatocellular cancer. Functional changes included resistance to acetaminophen and altered nitrogen metabolism. The transcriptomic landscape was characterized by two large clusters of monotonously increasing or decreasing genes, and a smaller number of ‘rest-and-jump genes’ that initially remained unaltered but became differentially expressed only at week 12 or later. Approximately 30% of the genes altered in human NAFLD are also altered in the present mouse model and an increasing overlap with genes altered in human HCC occurred at weeks 30–48. In conclusion, the observed sequence of events recapitulates many features of human disease and offers a basis for the identification of therapeutic targets.

Abstract Background Disease progression models are important for understanding the critical steps during the development of diseases. The models are imbedded in a statistical framework to deal with random variations due to biology and the sampling process when observing only a finite population. Conditional probabilities are used to describe dependencies between events that characterise the critical steps in the disease process. Many different model classes have been proposed in the literature, from simple path models to complex Bayesian networks. A popular and easy to understand but yet flexible model class are oncogenetic trees. These have been applied to describe the accumulation of genetic aberrations in cancer and HIV data. However, the number of potentially relevant aberrations is often by far larger than the maximal number of events that can be used for reliably estimating the progression models. Still, there are only a few approaches to variable selection, which have not yet been investigated in detail. Results We fill this gap and propose specifically for oncogenetic trees ten variable selection methods, some of these being completely new. We compare them in an extensive simulation study and on real data from cancer and HIV. It turns out that the preselection of events by clique identification algorithms performs best. Here, events are selected if they belong to the largest or the maximum weight subgraph in which all pairs of vertices are connected. Conclusions The variable selection method of identifying cliques finds both the important frequent events and those related to disease pathways.

Empirical analysis of statistical algorithms often demands time-consuming experiments. We present two R packages which greatly simplify working in batch computing environments. The package BatchJobs implements the basic objects and procedures to control any batch cluster from within R. It is structured around cluster versions of the well-known higher order functions Map, Reduce and Filter from functional programming. Computations are performed asynchronously and all job states are persistently stored in a database, which can be queried at any point in time. The second package, BatchExperiments, is tailored for the still very general scenario of analyzing arbitrary algorithms on problem instances. It extends package BatchJobs by letting the user define an array of jobs of the kind apply algorithm A to problem instance P and store results. It is possible to associate statistical designs with parameters of problems and algorithms and therefore to systematically study their influence on the results. The packages main features are: (a) Convenient usage: All relevant batch system operations are either handled internally or mapped to simple R functions. (b) Portability: Both packages use a clear and well-defined interface to the batch system which makes them applicable in most high-performance computing environments. (c) Reproducibility: Every computational part has an associated seed to ensure reproducibility even when the underlying batch system changes. (d) Abstraction and good software design: The code layers for algorithms, experiment definitions and execution are cleanly separated and enable the writing of readable and maintainable code.

Non-small cell lung cancer (NSCLC) represents a genomically unstable cancer type with extensive copy number aberrations. The relationship of gene copy number alterations and subsequent mRNA levels has only fragmentarily been described. The aim of this study was to conduct a genome-wide analysis of gene copy number gains and corresponding gene expression levels in a clinically well annotated NSCLC patient cohort (n = 190) and their association with survival. While more than half of all analyzed gene copy number-gene expression pairs showed statistically significant correlations (10,296 of 18,756 genes), high correlations, with a correlation coefficient >0.7, were obtained only in a subset of 301 genes (1.6%), including KRAS, EGFR and MDM2. Higher correlation coefficients were associated with higher copy number and expression levels. Strong correlations were frequently based on few tumors with high copy number gains and correspondingly increased mRNA expression. Among the highly correlating genes, GO groups associated with posttranslational protein modifications were particularly frequent, including ubiquitination and neddylation. In a meta-analysis including 1,779 patients we found that survival associated genes were overrepresented among highly correlating genes (61 of the 301 highly correlating genes, FDR adjusted p

Shallow marine ecosystems naturally experience fluctuating physicochemical conditions across spatial and temporal scales. Widespread coral-bleaching events, induced by prolonged heat stress, highlight the importance of how the duration and frequency of thermal stress influence the adaptive physiology of photosymbiotic calcifiers. Large benthic foraminifera harboring algal endosymbionts are major tropical carbonate producers and bioindicators of ecosystem health. Like corals, they are sensitive to thermal stress and bleach at temperatures temporarily occurring in their natural habitat and projected to happen more frequently. However, their thermal tolerance has been studied so far only by chronic exposure, so how they respond under more realistic episodic heat-event scenarios remains unknown. Here, we determined the physiological responses of Amphistegina gibbosa, an abundant western Atlantic foraminifera, to four different treatments--control, single, episodic, and chronic exposure to the same thermal stress (32°C)--in controlled laboratory cultures. Exposure to chronic thermal stress reduced motility and growth, while antioxidant capacity was elevated, and photosymbiont variables (coloration, oxygen-production rates, chlorophyll a concentration) indicated extensive bleaching. In contrast, single- and episodic-stress treatments were associated with higher motility and growth, while photosymbiont variables remained stable. The effects of single and episodic heat events were similar, except for the presumable occurrence of reproduction, which seemed to be suppressed by both episodic and chronic stress. The otherwise different responses between treatments with thermal fluctuations and chronic stress indicate adaptation to thermal peaks, but not to chronic exposure expected to ensue when baseline temperatures are elevated by climate change. This firstly implies that marine habitats with a history of fluctuating thermal stress potentially support resilient physiological mechanisms among photosymbiotic organisms. Secondly, there seem to be temporal constraints related to heat events among coral reef environments and reinforces the importance of temporal fluctuations in stress exposure in global-change studies and projections.

Disease classification from molecular measurements typically requires an analysis pipeline from raw noisy measurements to final classification results. Multi capillary column-ion mobility spectrometry (MCC-IMS) is a promising technology for the detection of volatile organic compounds in the air of exhaled breath. From raw measurements, the peak regions representing the compounds have to be identified, quantified, and clustered across different experiments. Currently, several steps of this analysis process require manual intervention of human experts. Our goal is to identify a fully automatic pipeline that yields competitive disease classification results compared to an established but subjective and tedious semi-manual process.We combine a large number of modern methods for peak detection, peak clustering, and multivariate classification into analysis pipelines for raw MCC-IMS data. We evaluate all combinations on three different real datasets in an unbiased cross-validation setting. We determine which specific algorithmic combinations lead to high AUC values in disease classifications across the different medical application scenarios.The best fully automated analysis process achieves even better classification results than the established manual process. The best algorithms for the three analysis steps are (i) SGLTR (Savitzky-Golay Laplace-operator filter thresholding regions) and LM (Local Maxima) for automated peak identification, (ii) EM clustering (Expectation Maximization) and DBSCAN (Density-Based Spatial Clustering of Applications with Noise) for the clustering step and (iii) RF (Random Forest) for multivariate classification. Thus, automated methods can replace the manual steps in the analysis process to enable an unbiased high throughput use of the technology.

TGR5 (Gpbar1) is a G protein-coupled receptor responsive to bile acids (BAs), which is expressed in different non-parenchymal cells of the liver, including biliary epithelial cells, liver-resident macrophages, sinusoidal endothelial cells (LSECs), and activated hepatic stellate cells (HSCs). Mice with targeted deletion of TGR5 are more susceptible towards cholestatic liver injury induced by cholic acid-feeding and bile duct ligation, resulting in a reduced proliferative response and increased liver injury. Conjugated lithocholic acid (LCA) represents the most potent TGR5 BA ligand and LCA-feeding has been used as a model to rapidly induce severe cholestatic liver injury in mice. Thus, TGR5 knockout (KO) mice and wildtype (WT) littermates were fed a diet supplemented with 1% LCA for 84 h. Liver injury and gene expression changes induced by the LCA diet revealed an enrichment of pathways associated with inflammation, proliferation, and matrix remodeling. Knockout of TGR5 in mice caused upregulation of endothelin-1 (ET-1) expression in the livers. Analysis of TGR5-dependent ET-1 signaling in isolated LSECs and HSCs demonstrated that TGR5 activation reduces ET-1 expression and secretion from LSECs and triggers internalization of the ET-1 receptor in HSCs, dampening ET-1 responsiveness. Thus, we identified two independent mechanisms by which TGR5 inhibits ET-1 signaling and modulates portal pressure.

Microarray experiments allow the monitoring of expression levels for thousands of genes simultaneously. Based on data obtained from the co-hybridization of two mRNA samples, a frequent goal is to find out which genes are differentially expressed. For this purpose, we propose to estimate the distribution of popular test statistics by a mixture of normal distributions. These statistics are calculated for each gene separately. A Bayes classifier is then used to decide upon differential expression. The cut-off for the classifier is chosen according to the number of false positives and negatives when applied to realistic data generating models. In particular, we generate data from a mixture model and from an Empirical Bayes model. By comparing the numbers of false decisions for various test statistics in the context of the considered models, we investigate which of the statistics are particularly suitable with our approach.

In breast cancer, gene signatures that predict the risk of metastasis after surgical tumor resection are mainly indicative of early events. The purpose of this study was to identify genes linked to metastatic recurrence more than three years after surgery.Affymetrix HG U133A and Plus 2.0 array datasets with information on metastasis-free, disease-free or overall survival were accessed via public repositories. Time restricted Cox regression models were used to identify genes associated with metastasis during or after the first three years post-surgery (early- and late-type genes). A sequential validation study design, with two non-adjuvantly treated discovery cohorts (n = 409) and one validation cohort (n = 169) was applied and identified genes were further evaluated in tamoxifen-treated breast cancer patients (n = 923), as well as in patients with non-small cell lung (n = 1779), colon (n = 893) and ovarian (n = 922) cancer.Ten late- and 243 early-type genes were identified in adjuvantly untreated breast cancer. Adjustment to clinicopathological factors and an established proliferation-related signature markedly reduced the number of early-type genes to 16, whereas nine late-type genes still remained significant. These nine genes were associated with metastasis-free survival (MFS) also in a non-time restricted model, but not in the early period alone, stressing that their prognostic impact was primarily based on MFS more than three years after surgery. Four of the ten late-type genes, the ribosome-related factors EIF4B, RPL5, RPL3, and the tumor angiogenesis modifier EPN3 were significantly associated with MFS in the late period also in a meta-analysis of tamoxifen-treated breast cancer cohorts. In contrast, only one late-type gene (EPN3) showed consistent survival associations in more than one cohort in the other cancer types, being associated with worse outcome in two non-small cell lung cancer cohorts. No late-type gene was validated in ovarian and colon cancer.Ribosome-related genes were associated with decreased risk of late metastasis in both adjuvantly untreated and tamoxifen-treated breast cancer patients. In contrast, high expression of epsin (EPN3) was associated with increased risk of late metastasis. This is of clinical relevance considering the well-understood role of epsins in tumor angiogenesis and the ongoing development of epsin antagonizing therapies.

Meningiomas are tumors that arise from the coverings of the brain or spinal cord. 5% of the cases turn into malignant forms with aggressive clinical behavior and increased risk of tumor recurrence. One hundred and five patients with meningiomas were operated by open surgery. To investigate predictors of meningioma recurrence in total 124 samples of 105 patients were investigated by iFISH. Dual-probe hybridization was performed to access chromosomal alterations of chromosomes 1p-, 9p- and 22q. Additionally, methylation of TIMP3 and p16 was analyzed with MS-PCR. Of the 105 investigated tumors 59.1% (62/105) were WHO grade I, 33.3% (35/105) were WHO grade II and 7.7% (8/105) were anaplastic meningiomas (grade III), respectively. The histopathological data correlates with the recurrence rate of the investigated meningiomas. Hypermethylation of TIMP3 was detected in 13.3% of all meningiomas: 10.9% in WHO grade I meningiomas, 25.0% in grade II and 14.3% in grade III meningiomas, respectively. No correlation of TIMP3 hypermethylation with tumor recurrence or WHO grade (p = 0.2) was observed. Interestingly, deletion of 1p36 emerged as a significant predictor of shorter overall survival (log rank test, p

Identifying relevant signatures for clinical patient outcome is a fundamental task in high-throughput studies. Signatures, composed of features such as mRNAs, miRNAs, SNPs or other molecular variables, are often non-overlapping, even though they have been identified from similar experiments considering samples with the same type of disease. The lack of a consensus is mostly due to the fact that sample sizes are far smaller than the numbers of candidate features to be considered, and therefore signature selection suffers from large variation. We propose a robust signature selection method that enhances the selection stability of penalized regression algorithms for predicting survival risk. Our method is based on an aggregation of multiple, possibly unstable, signatures obtained with the preconditioned lasso algorithm applied to random (internal) subsamples of a given cohort data, where the aggregated signature is shrunken by a simple thresholding strategy. The resulting method, RS-PL, is conceptually simple and easy to apply, relying on parameters automatically tuned by cross validation. Robust signature selection using RS-PL operates within an (external) subsampling framework to estimate the selection probabilities of features in multiple trials of RS-PL. These probabilities are used for identifying reliable features to be included in a signature. Our method was evaluated on microarray data sets from neuroblastoma, lung adenocarcinoma, and breast cancer patients, extracting robust and relevant signatures for predicting survival risk. Signatures obtained by our method achieved high prediction performance and robustness, consistently over the three data sets. Genes with high selection probability in our robust signatures have been reported as cancer-relevant. The ordering of predictor coefficients associated with signatures was well-preserved across multiple trials of RS-PL, demonstrating the capability of our method for identifying a transferable consensus signature. The software is available as an R package rsig at CRAN (http://cran.r-project.org).

The importance of inflammation is increasingly noticed in cancer. The aim of this study was to analyze the prognostic influence of pre-operative serum C-reactive protein (CRP) in a cohort of 148 lymph node-negative breast cancer patients. The prognostic significance of CRP level for disease-free survival (DFS), metastasis-free survival (MFS) and overall survival (OS) was evaluated using univariate and multivariate Cox regression, also including information on age at diagnosis, tumor size, tumor grade, estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) status, proliferation index (Ki67) and molecular subtype, as well as an assessment of the presence of necrosis and inflammation in the tumor tissue. Univariate analysis showed that CRP, as a continuous variable, was significantly associated with DFS (P = 0.002, hazard ratio [HR] = 1.04, 95% confidence interval [CI] = 1.02-1.07) and OS (P = 0.036, HR= 1.03, 95% CI = 1.00-1.06), whereas a trend was observed for MFS (P = 0.111). In the multivariate analysis, CRP retained its significance for DFS (P = 0.033, HR= 1.01, 95% CI = 1.00-1.07) as well as OS (P = 0.023, HR= 1.03, 95% CI = 1.00-1.06), independent of established prognostic factors. Furthermore, large-scale gene expression analysis by Affymetrix HG-U133A arrays was performed for 72 (48.6%) patients. The correlations between serum CRP and gene expression levels in the corresponding carcinoma of the breast were assessed using Spearman's rank correlation, controlled for false-discovery rate. No significant correlation was observed between CRP level and gene expression indicative of an ongoing local inflammatory process. In summary, pre-operatively elevated CRP levels at the time of diagnosis were associated with shorter DFS and OS independent of established prognostic factors in node-negative breast cancer, supporting a possible link between inflammation and prognosis in breast cancer.

Detection of yet unknown subgroups showing differential gene or protein expression is a frequent goal in the analysis of modern molecular data. Applications range from cancer biology over developmental biology to toxicology. Often a control and an experimental group are compared, and subgroups can be characterized by differential expression for only a subgroup-specific set of genes or proteins. Finding such genes and corresponding patient subgroups can help in understanding pathological pathways, diagnosis and defining drug targets. The size of the subgroup and the type of differential expression determine the optimal strategy for subgroup identification. To date, commonly used software packages hardly provide statistical tests and methods for the detection of such subgroups. Different univariate methods for subgroup detection are characterized and compared, both on simulated and on real data. We present an advanced design for simulation studies: Data is simulated under different distributional assumptions for the expression of the subgroup, and performance results are compared against theoretical upper bounds. For each distribution, different degrees of deviation from the majority of observations are considered for the subgroup. We evaluate classical approaches as well as various new suggestions in the context of omics data, including outlier sum, PADGE, and kurtosis. We also propose the new FisherSum score. ROC curve analysis and AUC values are used to quantify the ability of the methods to distinguish between genes or proteins with and without certain subgroup patterns. In general, FisherSum for small subgroups and t-test for large subgroups achieve best results. We apply each method to a case-control study on Parkinson's disease and underline the biological benefit of the new method.

Quantitative predictions in computational life sciences are often based on regression models. The advent of machine learning has led to highly accurate regression models that have gained widespread acceptance. While there are statistical methods available to estimate the global performance of regression models on a test or training dataset, it is often not clear how well this performance transfers to other datasets or how reliable an individual prediction is-a fact that often reduces a user's trust into a computational method. In analogy to the concept of an experimental error, we sketch how estimators for individual prediction errors can be used to provide confidence intervals for individual predictions. Two novel statistical methods, named CONFINE and CONFIVE, can estimate the reliability of an individual prediction based on the local properties of nearby training data. The methods can be applied equally to linear and non-linear regression methods with very little computational overhead. We compare our confidence estimators with other existing confidence and applicability domain estimators on two biologically relevant problems (MHC-peptide binding prediction and quantitative structure-activity relationship (QSAR)). Our results suggest that the proposed confidence estimators perform comparable to or better than previously proposed estimation methods. Given a sufficient amount of training data, the estimators exhibit error estimates of high quality. In addition, we observed that the quality of estimated confidence intervals is predictable. We discuss how confidence estimation is influenced by noise, the number of features, and the dataset size. Estimating the confidence in individual prediction in terms of error intervals represents an important step from plain, non-informative predictions towards transparent and interpretable predictions that will help to improve the acceptance of computational methods in the biological community.

Despite the successful introduction of potent anti-cancer therapeutics, most of these drugs lead to only modest tumor-shrinkage or transient responses, followed by re-growth of tumors. Combining different compounds has resulted in enhanced tumor control and prolonged survival. However, methods querying the efficacy of such combinations have been hampered by limited scalability, analytical resolution, statistical feasibility, or a combination thereof. We have developed a theoretical framework modeling cellular viability as a stochastic lifetime process to determine synergistic compound combinations from high-throughput cellular screens. We apply our method to data derived from chemical perturbations of 65 cancer cell lines with two inhibitors. Our analysis revealed synergy for the combination of both compounds in subsets of cell lines. By contrast, in cell lines in which inhibition of one of both targets was sufficient to induce cell death, no synergy was detected, compatible with the topology of the oncogenically activated signaling network. In summary, we provide a tool for the measurement of synergy strength for combination perturbation experiments that might help define pathway topologies and direct clinical trials.