Search

The process of Nuclei detection in high-grade breast cancer images is quite challenging in the case of image processing techniques due to certain heterogeneous characteristics of cancer nuclei such as enlarged and irregularly shaped nuclei, highly coarse chromatin marginalized to the nuclei periphery and visible nucleoli. Recent reviews state that existing techniques show appreciable segmentation accuracy on breast histopathology images whose nuclei are dispersed and regular in texture and shape; however, typical cancer nuclei are often clustered and have irregular texture and shape properties. This paper proposes a novel segmentation algorithm for detecting individual nuclei from Hematoxylin and Eosin (H&E) stained breast histopathology images. This detection framework estimates a nuclei saliency map using tensor voting followed by boundary extraction of the nuclei on the saliency map using a Loopy Back Propagation (LBP) algorithm on a Markov Random Field (MRF). The method was tested on both whole-slide images and frames of breast cancer histopathology images. Experimental results demonstrate high segmentation performance with efficient precision, recall and dice-coefficient rates, upon testing high-grade breast cancer images containing several thousand nuclei. In addition to the optimal performance on the highly complex images presented in this paper, this method also gave appreciable results in comparison with two recently published methods—Wienert et al. (2012) and Veta et al. (2013), which were tested using their own datasets. [ABSTRACT FROM AUTHOR]

Feature selection and classification are the main topics in microarray data analysis. Although many feature selection methods have been proposed and developed in this field, SVM-RFE (Support Vector Machine based on Recursive Feature Elimination) is proved as one of the best feature selection methods, which ranks the features (genes) by training support vector machine classification model and selects key genes combining with recursive feature elimination strategy. The principal drawback of SVM-RFE is the huge time consumption. To overcome this limitation, we introduce a more efficient implementation of linear support vector machines and improve the recursive feature elimination strategy and then combine them together to select informative genes. Besides, we propose a simple resampling method to preprocess the datasets, which makes the information distribution of different kinds of samples balanced and the classification results more credible. Moreover, the applicability of four common classifiers is also studied in this paper. Extensive experiments are conducted on six most frequently used microarray datasets in this field, and the results show that the proposed methods have not only reduced the time consumption greatly but also obtained comparable classification performance. [ABSTRACT FROM AUTHOR]

Recent research has demonstrated that characteristic gene selection based on gene expression data remains faced with considerable challenges. This is primarily because gene expression data are typically high dimensional, negative, non-sparse and noisy. However, existing methods for data analysis are able to cope with only some of these challenges. In this paper, we address all of these challenges with a unified method: nonnegative matrix factorization via the L2,1-norm (NMF-L2,1). While L2,1-norm minimization is applied to both the error function and the regularization term, our method is robust to outliers and noise in the data and generates sparse results. The application of our method to plant and tumor gene expression data demonstrates that NMF-L2,1 can extract more characteristic genes than other existing state-of-the-art methods. [ABSTRACT FROM AUTHOR]

The identification of a subset of genes having the ability to capture the necessary information to distinguish classes of patients is crucial in bioinformatics applications. Ensemble and bagging methods have been shown to work effectively in the process of gene selection and classification. Testament to that is random forest which combines random decision trees with bagging to improve overall feature selection and classification accuracy. Surprisingly, the adoption of these methods in support vector machines has only recently received attention but mostly on classification not gene selection. This paper introduces an ensemble SVM-Recursive Feature Elimination (ESVM-RFE) for gene selection that follows the concepts of ensemble and bagging used in random forest but adopts the backward elimination strategy which is the rationale of RFE algorithm. The rationale behind this is, building ensemble SVM models using randomly drawn bootstrap samples from the training set, will produce different feature rankings which will be subsequently aggregated as one feature ranking. As a result, the decision for elimination of features is based upon the ranking of multiple SVM models instead of choosing one particular model. Moreover, this approach will address the problem of imbalanced datasets by constructing a nearly balanced bootstrap sample. Our experiments show that ESVM-RFE for gene selection substantially increased the classification performance on five microarray datasets compared to state-of-the-art methods. Experiments on the childhood leukaemia dataset show that an average 9% better accuracy is achieved by ESVM-RFE over SVM-RFE, and 5% over random forest based approach. The selected genes by the ESVM-RFE algorithm were further explored with Singular Value Decomposition (SVD) which reveals significant clusters with the selected data. [ABSTRACT FROM AUTHOR]

Cancer classification and feature (gene) selection plays an important role in knowledge discovery in genomic data. Although logistic regression is one of the most popular classification methods, it does not induce feature selection. In this paper, we presented a new hybrid L1/2 +2 regularization (HLR) function, a linear combination of L1/2 and L2 penalties, to select the relevant gene in the logistic regression. The HLR approach inherits some fascinating characteristics from L1/2 (sparsity) and L2 (grouping effect where highly correlated variables are in or out a model together) penalties. We also proposed a novel univariate HLR thresholding approach to update the estimated coefficients and developed the coordinate descent algorithm for the HLR penalized logistic regression model. The empirical results and simulations indicate that the proposed method is highly competitive amongst several state-of-the-art methods. [ABSTRACT FROM AUTHOR]

For studying cancer and genetic diseases, the issue of identifying high correlation genes from high-dimensional data is an important problem. It is a great challenge to select relevant biomarkers from gene expression data that contains some important correlation structures, and some of the genes can be divided into different groups with a common biological function, chromosomal location or regulation. In this paper, we propose a penalized accelerated failure time model CHR-DE using a non-convex regularization (local search) with differential evolution (global search) in a wrapper-embedded memetic framework. The complex harmonic regularization (CHR) can approximate to the combination and ℓq (1 ≤ q < 2) for selecting biomarkers in group. And differential evolution (DE) is utilized to globally optimize the CHR’s hyperparameters, which make CHR-DE achieve strong capability of selecting groups of genes in high-dimensional biological data. We also developed an efficient path seeking algorithm to optimize this penalized model. The proposed method is evaluated on synthetic and three gene expression datasets: breast cancer, hepatocellular carcinoma and colorectal cancer. The experimental results demonstrate that CHR-DE is a more effective tool for feature selection and learning prediction. [ABSTRACT FROM AUTHOR]

Several modern genomic technologies, such as DNA-Methylation arrays, measure spatially registered probes that number in the hundreds of thousands across multiple chromosomes. The measured probes are by themselves less interesting scientifically; instead scientists seek to discover biologically interpretable genomic regions comprised of contiguous groups of probes which may act as biomarkers of disease or serve as a dimension-reducing pre-processing step for downstream analyses. In this paper, we introduce an unsupervised feature learning technique which maps technological units (probes) to biological units (genomic regions) that are common across all subjects. We use ideas from fusion penalties and convex clustering to introduce a method for Spatial Convex Clustering, or SpaCC. Our method is specifically tailored to detecting multi-subject regions of methylation, but we also test our approach on the well-studied problem of detecting segments of copy number variation. We formulate our method as a convex optimization problem, develop a massively parallelizable algorithm to find its solution, and introduce automated approaches for handling missing values and determining tuning parameters. Through simulation studies based on real methylation and copy number variation data, we show that SpaCC exhibits significant performance gains relative to existing methods. Finally, we illustrate SpaCC’s advantages as a pre-processing technique that reduces large-scale genomics data into a smaller number of genomic regions through several cancer epigenetics case studies on subtype discovery, network estimation, and epigenetic-wide association. [ABSTRACT FROM AUTHOR]

Microarray technology is important to simultaneously express multiple genes over a number of time points. Multiple classifier models, such as sparse representation (SR)-based method, have been developed to classify microarray gene expression data. These methods allocate the gene data points to different clusters. In this paper, we propose a novel collaborative representation (CR)-based classification with regularized least square to classify gene data. First, the CR codes a testing sample as a sparse linear combination of all training samples and then classifies the testing sample by evaluating which class leads to the minimum representation error. This CR-based classification approach is remarkably less complex than traditional classification methods but leads to very competitive classification results. In addition, compressive sensing approach is adopted to project the high-dimensional gene expression dataset to a lower-dimensional space which nearly contains the whole information. This compression without loss is beneficial to reduce the computational load. Experiments to detect subtypes of diseases, such as leukemia and autism spectrum disorders, are performed by analyzing the gene expression. The results show that the proposed CR-based algorithm exhibits significantly higher stability and accuracy than the traditional classifiers, such as support vector machine algorithm. [ABSTRACT FROM AUTHOR]

In microarray data analysis, we are often required to combine several dependent partial test results. To overcome this, many suggestions have been made in previous literature; Tippett’s test and Fisher’s omnibus test are most popular. Both tests have known null distributions when the partial tests are independent. However, for dependent tests, their (even, asymptotic) null distributions are unknown and additional numerical procedures are required. In this paper, we revisited Stouffer’s test based on z-scores and showed its advantage over the two aforementioned methods in the analysis of large-scale microarray data. The combined statistic in Stouffer’s test has a normal distribution with mean 0 from the normality of the z-scores. Its variance can be estimated from the scores of genes in the experiment without an additional numerical procedure. We numerically compared the errors of Stouffer’s test and the two p-value based methods, Tippett’s test and Fisher’s omnibus test. We also analyzed our microarray data to find differentially expressed genes by non-genotoxic and genotoxic carcinogen compounds. Both numerical study and the real application showed that Stouffer’s test performed better than Tippett’s method and Fisher’s omnibus method with additional permutation steps. [ABSTRACT FROM AUTHOR]

Integrative clustering is used to identify groups of samples by jointly analysing multiple datasets describing the same set of biological samples, such as gene expression, copy number, methylation etc. Most existing algorithms for integrative clustering assume that there is a shared consistent set of clusters across all datasets, and most of the data samples follow this structure. However in practice, the structure across heterogeneous datasets can be more varied, with clusters being joined in some datasets and separated in others. In this paper, we present a probabilistic clustering method to identify groups across datasets that do not share the same cluster structure. The proposed algorithm, Clusternomics, identifies groups of samples that share their global behaviour across heterogeneous datasets. The algorithm models clusters on the level of individual datasets, while also extracting global structure that arises from the local cluster assignments. Clusters on both the local and the global level are modelled using a hierarchical Dirichlet mixture model to identify structure on both levels. We evaluated the model both on simulated and on real-world datasets. The simulated data exemplifies datasets with varying degrees of common structure. In such a setting Clusternomics outperforms existing algorithms for integrative and consensus clustering. In a real-world application, we used the algorithm for cancer subtyping, identifying subtypes of cancer from heterogeneous datasets. We applied the algorithm to TCGA breast cancer dataset, integrating gene expression, miRNA expression, DNA methylation and proteomics. The algorithm extracted clinically meaningful clusters with significantly different survival probabilities. We also evaluated the algorithm on lung and kidney cancer TCGA datasets with high dimensionality, again showing clinically significant results and scalability of the algorithm. [ABSTRACT FROM AUTHOR]

Gene expression data analysis has paramount implications for gene treatments, cancer diagnosis and other domains. Clustering is an important and promising tool to analyze gene expression data. Gene expression data is often characterized by a large amount of genes but with limited samples, thus various projective clustering techniques and ensemble techniques have been suggested to combat with these challenges. However, it is rather challenging to synergy these two kinds of techniques together to avoid the curse of dimensionality problem and to boost the performance of gene expression data clustering. In this paper, we employ a projective clustering ensemble (PCE) to integrate the advantages of projective clustering and ensemble clustering, and to avoid the dilemma of combining multiple projective clusterings. Our experimental results on publicly available cancer gene expression data show PCE can improve the quality of clustering gene expression data by at least 4.5% (on average) than other related techniques, including dimensionality reduction based single clustering and ensemble approaches. The empirical study demonstrates that, to further boost the performance of clustering cancer gene expression data, it is necessary and promising to synergy projective clustering with ensemble clustering. PCE can serve as an effective alternative technique for clustering gene expression data. [ABSTRACT FROM AUTHOR]

One of fundamental challenges in cancer studies is that varying molecular characteristics of different tumor types may lead to resistance to certain drugs. As a result, the same drug can lead to significantly different results in different types of cancer thus emphasizing the need for individualized medicine. Individual prediction of drug response has great potential to aid in improving the clinical outcome and reduce the financial costs associated with prescribing chemotherapy drugs to which the patient’s tumor might be resistant. In this paper we develop a network based classifier (NBC) method for predicting sensitivity of cell lines to anticancer drugs from transcriptome data. In the literature, this strategy has been used for predicting cancer types. Here, we extend it to estimate sensitivity of cells from different tumor types to various anticancer drugs. Furthermore, we incorporate domain specific knowledge such as the use of apoptotic gene list and clinical dose information in our method to impart biological significance to the prediction. Our experimental results suggest that our network based classifier (NBC) method outperforms existing classifiers in estimating sensitivity of cell lines for different drugs. [ABSTRACT FROM AUTHOR]

Disease module is a group of molecular components that interact intensively in the disease specific biological network. Since the connectivity and activity of disease modules may shed light on the molecular mechanisms of pathogenesis and disease progression, their identification becomes one of the most important challenges in network medicine, an emerging paradigm to study complex human disease. This paper proposes a novel algorithm, DiME (Disease Module Extraction), to identify putative disease modules from biological networks. We have developed novel heuristics to optimise Community Extraction, a module criterion originally proposed for social network analysis, to extract topological core modules from biological networks as putative disease modules. In addition, we have incorporated a statistical significance measure, B-score, to evaluate the quality of extracted modules. As an application to complex diseases, we have employed DiME to investigate the molecular mechanisms that underpin the progression of glioma, the most common type of brain tumour. We have built low (grade II) - and high (GBM) - grade glioma co-expression networks from three independent datasets and then applied DiME to extract potential disease modules from both networks for comparison. Examination of the interconnectivity of the identified modules have revealed changes in topology and module activity (expression) between low- and high- grade tumours, which are characteristic of the major shifts in the constitution and physiology of tumour cells during glioma progression. Our results suggest that transcription factors E2F4, AR and ETS1 are potential key regulators in tumour progression. Our DiME compiled software, R/C++ source code, sample data and a tutorial are available at http://www.cs.bham.ac.uk/~szh/DiME. [ABSTRACT FROM AUTHOR]

miRNAs are small non-coding RNAs that regulate gene expression by binding to the 3′-UTR of genes. Many recent studies have reported that miRNAs play important biological roles by regulating specific mRNAs or genes. Many sequence-based target prediction algorithms have been developed to predict miRNA targets. However, these methods are not designed for condition-specific target predictions and produce many false positives; thus, expression-based target prediction algorithms have been developed for condition-specific target predictions. A typical strategy to utilize expression data is to leverage the negative control roles of miRNAs on genes. To control false positives, a stringent cutoff value is typically set, but in this case, these methods tend to reject many true target relationships, i.e., false negatives. To overcome these limitations, additional information should be utilized. The literature is probably the best resource that we can utilize. Recent literature mining systems compile millions of articles with experiments designed for specific biological questions, and the systems provide a function to search for specific information. To utilize the literature information, we used a literature mining system, BEST, that automatically extracts information from the literature in PubMed and that allows the user to perform searches of the literature with any English words. By integrating omics data analysis methods and BEST, we developed Context-MMIA, a miRNA-mRNA target prediction method that combines expression data analysis results and the literature information extracted based on the user-specified context. In the pathway enrichment analysis using genes included in the top 200 miRNA-targets, Context-MMIA outperformed the four existing target prediction methods that we tested. In another test on whether prediction methods can re-produce experimentally validated target relationships, Context-MMIA outperformed the four existing target prediction methods. In summary, Context-MMIA allows the user to specify a context of the experimental data to predict miRNA targets, and we believe that Context-MMIA is very useful for predicting condition-specific miRNA targets. [ABSTRACT FROM AUTHOR]

In transcriptomics, micro RNAs (miRNAs) has gained much interest especially as potential disease indicators. However, apart from holding a great promise related to their clinical application, a lot of inconsistent results have been published. Our aim was to compare the miRNA expression levels in ovarian cancer and healthy subjects using the Bayesian multilevel model and to assess their potential usefulness in diagnosis. We have analyzed a case-control observational data on expression profiling of 49 preselected miRNA-based ovarian cancer indicators in 119 controls and 59 patients. A Bayesian multilevel model was used to characterize the effect of disease on miRNA levels controlling for differences in age and body weight. The difference between the miRNA level and health status of the patient on the scale of the data variability were discussed in the context of their potential usefulness in diagnosis. Additionally, the cross-validated area under the ROC curve (AUC) was used to assess the expected out-of-sample discrimination index of a different sets of miRNAs. The proposed model allowed us to describe the set of miRNA levels in patients and controls. Three highly correlated miRNAs: miR-101-3p, miR-142-5p, miR-148a-3p rank the highest with almost identical effect sizes that ranges from 0.45 to 1.0. For those miRNAs the credible interval for AUC ranged from 0.63 to 0.67 indicating their limited discrimination potential. A little benefit in adding information from other miRNAs was observed. There were several miRNAs in the dataset (miR-604, hsa-miR-221-5p) for which inferences were uncertain. For those miRNAs more experimental effort is needed to fully assess their effect in the context of new hits discovery and usefulness as disease indicators. The proposed multilevel Bayesian model can be used to characterize the panel of miRNA profile and to assess the difference in expression levels between healthy and cancer individuals. [ABSTRACT FROM AUTHOR]

Background: Gene shaving (GS) is an essential and challenging tools for biomedical researchers due to the large number of genes in human genome and the complex nature of biological networks. Most GS methods are not applicable to non-linear and multi-view data sets. While the kernel based methods can overcome these problems, a well-founded positive definite kernel based GS method has yet to be proposed for biomedical data analysis. Methods and findings: Since the kernel based methods on genomic information can improve the prediction of diseases, here we proposed a noble method, “kernel based gene shaving” which is based on the influence function of kernel canonical correlation analysis. To investigate the performance of the proposed method in comparison to state-of-the-art-method in gene saving, we analyzed extensive simulated and real microarray gene expression data set. The performance metrics including true positive rate, true negative rate, false positive rate, false negative rate, misclassification error rate, the false discovery rate and area under curves were computed for each methods. In colon cancer data analysis, the proposed method identified a significant subsets of 210 genes out of 2000 genes and suggestive superior performance compared with other methods. The proposed method can be applied to the study of other disease process where two view data is a common task. Conclusions: We addressed the challenge of finding unique kernel based GS methods by using the influence function of kernel canonical correlation analysis. The proposed method has shown to have better performance than state-of-the-art-methods in gene saving and has identified many more significant gene interactions, suggesting that genes function in a concerted effort in colon cancer. In similar biomedical data analysis, kernel based methods could be applied to select a potential subset of genes. The positive definite kernel based methods can overcome the non-linearity problem and improve the prediction process. [ABSTRACT FROM AUTHOR]

MiRNAs and proteins play important roles in different stages of breast tumor development and serve as biomarkers for the early diagnosis of breast cancer. A new algorithm that combines machine learning algorithms and multilayer complex network analysis is hereby proposed to explore the potential diagnostic values of miRNAs and proteins. XGBoost and random forest algorithms were employed to screen the most important miRNAs and proteins. Maximal information coefficient was applied to assess intralayer and interlayer connection. A multilayer complex network was constructed to identify miRNAs and proteins that could serve as biomarkers for breast cancer. Proteins and miRNAs that are nodes in the network were subsequently categorized into two network layers considering their distinct functions. The betweenness centrality was used as the first measurement of the importance of the nodes within each single layer. The degree of the nodes was chosen as the second measurement to map their signalling pathways. By combining these two measurements into one score and comparing the difference of the same candidate between normal tissue and cancer tissue, this novel multilayer network analysis could be applied to successfully identify molecules associated with breast cancer. [ABSTRACT FROM AUTHOR]

Increasing evidence has indicated that microRNAs(miRNAs) play vital roles in various pathological processes and thus are closely related with many complex human diseases. The identification of potential disease-related miRNAs offers new opportunities to understand disease etiology and pathogenesis. Although there have been numerous computational methods proposed to predict reliable miRNA-disease associations, they suffer from various limitations that affect the prediction accuracy and their applicability. In this study, we develop a novel method to discover disease-related candidate miRNAs based on Adaptive Multi-View Multi-Label learning(AMVML). Specifically, considering the inherent noise existed in the current dataset, we propose to learn a new affinity graph adaptively for both diseases and miRNAs from multiple similarity profiles. We then simultaneously update the miRNA-disease association predicted from both spaces based on multi-label learning. In particular, we prove the convergence of AMVML theoretically and the corresponding analysis indicates that it has a fast convergence rate. To comprehensively illustrate the prediction performance of our method, we compared AMVML with four state-of-the-art methods under different validation frameworks. As a result, our method achieves comparable performance under various evaluation metrics, which suggests that our method is capable of discovering greater number of true miRNA-disease associations. The case study conducted on thyroid neoplasms further identified a potential diagnostic biomarker. Together, the experimental results confirms the utility of our method and we anticipate that our method could serve as a reliable and efficient tool for uncovering novel disease-related miRNAs. [ABSTRACT FROM AUTHOR]

Since their inception, several tools have been developed for cluster analysis and heatmap construction. The application of such tools to the number and types of genome-wide data available from next generation sequencing (NGS) technologies requires the adaptation of statistical concepts, such as in defining a most variable gene set, and more intricate cluster analyses method to address multiple omic data types. Additionally, the growing number of publicly available datasets has created the desire to estimate the statistical significance of a gene signature derived from one dataset to similarly group samples based on another dataset. The currently available number of tools and their combined use for generating heatmaps, along with the several adaptations of statistical concepts for addressing the higher dimensionality of genome-wide NGS-derived data, has created a further challenge in the ability to replicate heatmap results. We introduce NOJAH (NOt Just Another Heatmap), an interactive tool that defines and implements a workflow for genome-wide cluster analysis and heatmap construction by creating and combining several tools into a single user interface. NOJAH includes several newly developed scripts for techniques that though frequently applied are not sufficiently documented to allow for replicability of results. These techniques include: defining a most variable gene set (a.k.a., ‘core genes’), estimating the statistical significance of a gene signature to separate samples into clusters, and performing a result merging integrated cluster analysis. With only a user uploaded dataset, NOJAH provides as output, among other things, the minimum documentation required for replicating heatmap results. Additionally, NOJAH contains five different existing R packages that are connected in the interface by their functionality as part of a defined workflow for genome-wide cluster analysis. The NOJAH application tool is available at with corresponding source code available at . [ABSTRACT FROM AUTHOR]

In more recent years, a significant increase in the number of available biological experiments has taken place due to the widespread use of massive sequencing data. Furthermore, the continuous developments in the machine learning and in the high performance computing areas, are allowing a faster and more efficient analysis and processing of this type of data. However, biological information about a certain disease is normally widespread due to the use of different sequencing technologies and different manufacturers, in different experiments along the years around the world. Thus, nowadays it is of paramount importance to attain a correct integration of biologically-related data in order to achieve genuine benefits from them. For this purpose, this work presents an integration of multiple Microarray and RNA-seq platforms, which has led to the design of a multiclass study by collecting samples from the main four types of leukemia, quantified at gene expression. Subsequently, in order to find a set of differentially expressed genes with the highest discernment capability among different types of leukemia, an innovative parameter referred to as coverage is presented here. This parameter allows assessing the number of different pathologies that a certain gen is able to discern. It has been evaluated together with other widely known parameters under assessment of an ANOVA statistical test which corroborated its filtering power when the identified genes are subjected to a machine learning process at multiclass level. The optimal tuning of gene extraction evaluated parameters by means of this statistical test led to the selection of 42 highly relevant expressed genes. By the use of minimum-Redundancy Maximum-Relevance (mRMR) feature selection algorithm, these genes were reordered and assessed under the operation of four different classification techniques. Outstanding results were achieved by taking exclusively the first ten genes of the ranking into consideration. Finally, specific literature was consulted on this last subset of genes, revealing the occurrence of practically all of them with biological processes related to leukemia. At sight of these results, this study underlines the relevance of considering a new parameter which facilitates the identification of highly valid expressed genes for simultaneously discerning multiple types of leukemia. [ABSTRACT FROM AUTHOR]

Chromosomal aberrations, such as micronuclei (MN), have served as biomarkers of genotoxic exposure and cancer risk. Guidelines for the process of scoring MN have been presented by the HUman MicroNucleus (HUMN) project. However, these guidelines were developed for assay performance but do not address how to statistically analyze the data generated by the assay. This has led to the application of various statistical methods that may render different interpretations and conclusions. By combining MN with data from other high-throughput genomic technologies such as gene expression microarray data, we may elucidate molecular features involved in micronucleation. Traditional methods that can model discrete (synonymously, count) data, such as MN frequency, require that the number of explanatory variables (P) is less than the number of samples (N). Due to this limitation, penalized models have been developed to enable model fitting for such over-parameterized datasets. Because penalized models in the discrete response setting are lacking, particularly when the count outcome is over-dispersed, herein we present our penalized negative binomial regression model that can be fit when P > N. Using simulation studies we demonstrate the performance of our method in comparison to commonly used penalized Poisson models when the outcome is over-dispersed and applied it to MN frequency and gene expression data collected as part of the Norwegian Mother and Child Cohort Study. Our R package is available for download from the Comprehensive R Archive Network and can be applied to datasets having a discrete outcome that is either Poisson or negative binomial distributed and a high-dimensional covariate space. [ABSTRACT FROM AUTHOR]

We introduce the CDRP (Concatenated Diagnostic-Relapse Prognostic) architecture for multi-task deep learning that incorporates a clinical algorithm, e.g., a risk stratification schema to improve prognostic profiling. We present the first application to survival prediction in High-Risk (HR) Neuroblastoma from transcriptomics data, a task that studies from the MAQC consortium have shown to remain the hardest among multiple diagnostic and prognostic endpoints predictable from the same dataset. To obtain a more accurate risk stratification needed for appropriate treatment strategies, CDRP combines a first component (CDRP-A) synthesizing a diagnostic task and a second component (CDRP-N) dedicated to one or more prognostic tasks. The approach leverages the advent of semi-supervised deep learning structures that can flexibly integrate multimodal data or internally create multiple processing paths. CDRP-A is an autoencoder trained on gene expression on the HR/non-HR risk stratification by the Children’s Oncology Group, obtaining a 64-node representation in the bottleneck layer. CDRP-N is a multi-task classifier for two prognostic endpoints, i.e., Event-Free Survival (EFS) and Overall Survival (OS). CDRP-A provides the HR embedding input to the CDRP-N shared layer, from which two branches depart to model EFS and OS, respectively. To control for selection bias, CDRP is trained and evaluated using a Data Analysis Protocol (DAP) developed within the MAQC initiative. CDRP was applied on Illumina RNA-Seq of 498 Neuroblastoma patients (HR: 176) from the SEQC study (12,464 Entrez genes) and on Affymetrix Human Exon Array expression profiles (17,450 genes) of 247 primary diagnostic Neuroblastoma of the TARGET NBL cohort. On the SEQC HR patients, CDRP achieves Matthews Correlation Coefficient (MCC) 0.38 for EFS and MCC = 0.19 for OS in external validation, improving over published SEQC models. We show that a CDRP-N embedding is indeed parametrically associated to increasing severity and the embedding can be used to better stratify patients’ survival. [ABSTRACT FROM AUTHOR]

It is becoming increasingly clear that major malignancies such as breast, colorectal and gastric cancers are not single disease entities, but comprising multiple cancer subtypes of distinct molecular properties. Molecular subtyping has been widely used to dissect inter-tumor biological heterogeneity, in relation to clinical outcomes. A key step of this methodology is to perform unsupervised classification of gene expression profiles, which, however, often suffers challenges of high-dimensionality, feature redundancy as well as noise and irrelevant information. To overcome these limitations, we propose ELM-CC, which employs hidden observation features obtained from extreme learning machines (ELMs) for cancer classification. To demonstrate the effectiveness and usefulness, we applied ELM-CC for gastric and ovarian cancer subtyping. Comparing with the widely-used consensus clustering method, our approach demonstrated much better clustering performance and identified molecular subtypes that are much more clinically relevant. [ABSTRACT FROM AUTHOR]

Recently, a growing number of biological research and scientific experiments have demonstrated that microRNA (miRNA) affects the development of human complex diseases. Discovering miRNA-disease associations plays an increasingly vital role in devising diagnostic and therapeutic tools for diseases. However, since uncovering associations via experimental methods is expensive and time-consuming, novel and effective computational methods for association prediction are in demand. In this study, we developed a computational model of Matrix Decomposition and Heterogeneous Graph Inference for miRNA-disease association prediction (MDHGI) to discover new miRNA-disease associations by integrating the predicted association probability obtained from matrix decomposition through sparse learning method, the miRNA functional similarity, the disease semantic similarity, and the Gaussian interaction profile kernel similarity for diseases and miRNAs into a heterogeneous network. Compared with previous computational models based on heterogeneous networks, our model took full advantage of matrix decomposition before the construction of heterogeneous network, thereby improving the prediction accuracy. MDHGI obtained AUCs of 0.8945 and 0.8240 in the global and the local leave-one-out cross validation, respectively. Moreover, the AUC of 0.8794+/-0.0021 in 5-fold cross validation confirmed its stability of predictive performance. In addition, to further evaluate the model's accuracy, we applied MDHGI to four important human cancers in three different kinds of case studies. In the first type, 98% (Esophageal Neoplasms) and 98% (Lymphoma) of top 50 predicted miRNAs have been confirmed by at least one of the two databases (dbDEMC and miR2Disease) or at least one experimental literature in PubMed. In the second type of case study, what made a difference was that we removed all known associations between the miRNAs and Lung Neoplasms before implementing MDHGI on Lung Neoplasms. As a result, 100% (Lung Neoplasms) of top 50 related miRNAs have been indexed by at least one of the three databases (dbDEMC, miR2Disease and HMDD V2.0) or at least one experimental literature in PubMed. Furthermore, we also tested our prediction method on the HMDD V1.0 database to prove the applicability of MDHGI to different datasets. The results showed that 50 out of top 50 miRNAs related with the breast neoplasms were validated by at least one of the three databases (HMDD V2.0, dbDEMC, and miR2Disease) or at least one experimental literature. [ABSTRACT FROM AUTHOR]

MicroRNAs are small non-coding RNAs that influence gene expression by binding to the 3’ UTR of target mRNAs in order to repress protein synthesis. Soon after discovery, microRNA dysregulation has been associated to several pathologies. In particular, they have often been reported as differentially expressed in healthy and tumor samples. This fact suggested that microRNAs are likely to be good candidate biomarkers for cancer diagnosis and personalized medicine. With the advent of Next-Generation Sequencing (NGS), measuring the expression level of the whole miRNAome at once is now routine. Yet, the collaborative effort of sharing data opens to the possibility of population analyses. This context motivated us to perform an in-silico study to distill cancer-specific panels of microRNAs that can serve as biomarkers. We observed that the problem of finding biomarkers can be modeled as a two-class classification task where, given the miRNAomes of a population of healthy and cancerous samples, we want to find the subset of microRNAs that leads to the highest classification accuracy. We fulfill this task leveraging on a sensible combination of data mining tools. In particular, we used: differential evolution for candidate selection, component analysis to preserve the relationships among miRNAs, and SVM for sample classification. We identified 10 cancer-specific panels whose classification accuracy is always higher than 92%. These panels have a very little overlap suggesting that miRNAs are not only predictive of the onset of cancer, but can be used for classification purposes as well. We experimentally validated the contribution of each of the employed tools to the selection of discriminating miRNAs. Moreover, we tested the significance of each panel for the corresponding cancer type. In particular, enrichment analysis showed that the selected miRNAs are involved in oncogenesis pathways, while survival analysis proved that miRNAs can be used to evaluate cancer severity. Summarizing: results demonstrated that our method is able to produce cancer-specific panels that are promising candidates for a subsequent in vitro validation. [ABSTRACT FROM AUTHOR]

Analysis of data measured on different scales is a relevant challenge. Biomedical studies often focus on high-throughput datasets of, e.g., quantitative measurements. However, the need for integration of other features possibly measured on different scales, e.g. clinical or cytogenetic factors, becomes increasingly important. The analysis results (e.g. a selection of relevant genes) are then visualized, while adding further information, like clinical factors, on top. However, a more integrative approach is desirable, where all available data are analyzed jointly, and where also in the visualization different data sources are combined in a more natural way. Here we specifically target integrative visualization and present a heatmap-style graphic display. To this end, we develop and explore methods for clustering mixed-type data, with special focus on clustering variables. Clustering of variables does not receive as much attention in the literature as does clustering of samples. We extend the variables clustering methodology by two new approaches, one based on the combination of different association measures and the other on distance correlation. With simulation studies we evaluate and compare different clustering strategies. Applying specific methods for mixed-type data proves to be comparable and in many cases beneficial as compared to standard approaches applied to corresponding quantitative or binarized data. Our two novel approaches for mixed-type variables show similar or better performance than the existing methods ClustOfVar and bias-corrected mutual information. Further, in contrast to ClustOfVar, our methods provide dissimilarity matrices, which is an advantage, especially for the purpose of visualization. Real data examples aim to give an impression of various kinds of potential applications for the integrative heatmap and other graphical displays based on dissimilarity matrices. We demonstrate that the presented integrative heatmap provides more information than common data displays about the relationship among variables and samples. The described clustering and visualization methods are implemented in our R package CluMix available from . [ABSTRACT FROM AUTHOR]

Co-expression networks have long been used as a tool for investigating the molecular circuitry governing biological systems. However, most algorithms for constructing co-expression networks were developed in the microarray era, before high-throughput sequencing—with its unique statistical properties—became the norm for expression measurement. Here we develop Bayesian Relevance Networks, an algorithm that uses Bayesian reasoning about expression levels to account for the differing levels of uncertainty in expression measurements between highly- and lowly-expressed entities, and between samples with different sequencing depths. It combines data from groups of samples (e.g., replicates) to estimate group expression levels and confidence ranges. It then computes uncertainty-moderated estimates of cross-group correlations between entities, and uses permutation testing to assess their statistical significance. Using large scale miRNA data from The Cancer Genome Atlas, we show that our Bayesian update of the classical Relevance Networks algorithm provides improved reproducibility in co-expression estimates and lower false discovery rates in the resulting co-expression networks. Software is available at . [ABSTRACT FROM AUTHOR]

Integrative analyses of high-throughput ‘omic data, such as DNA methylation, DNA copy number alteration, mRNA and protein expression levels, have created unprecedented opportunities to understand the molecular basis of human disease. In particular, integrative analyses have been the cornerstone in the study of cancer to determine molecular subtypes within a given cancer. As malignant tumors with similar morphological characteristics have been shown to exhibit entirely different molecular profiles, there has been significant interest in using multiple ‘omic data for the identification of novel molecular subtypes of disease, which could impact treatment decisions. Therefore, we have developed intNMF, an integrative approach for disease subtype classification based on non-negative matrix factorization. The proposed approach carries out integrative clustering of multiple high dimensional molecular data in a single comprehensive analysis utilizing the information across multiple biological levels assessed on the same individual. As intNMF does not assume any distributional form for the data, it has obvious advantages over other model based clustering methods which require specific distributional assumptions. Application of intNMF is illustrated using both simulated and real data from The Cancer Genome Atlas (TCGA). [ABSTRACT FROM AUTHOR]

In the recent few years, an increasing number of studies have shown that microRNAs (miRNAs) play critical roles in many fundamental and important biological processes. As one of pathogenetic factors, the molecular mechanisms underlying human complex diseases still have not been completely understood from the perspective of miRNA. Predicting potential miRNA-disease associations makes important contributions to understanding the pathogenesis of diseases, developing new drugs, and formulating individualized diagnosis and treatment for diverse human complex diseases. Instead of only depending on expensive and time-consuming biological experiments, computational prediction models are effective by predicting potential miRNA-disease associations, prioritizing candidate miRNAs for the investigated diseases, and selecting those miRNAs with higher association probabilities for further experimental validation. In this study, ath-ased iRNA-isease ssociation (PBMDA) prediction model was proposed by integrating known human miRNA-disease associations, miRNA functional similarity, disease semantic similarity, and Gaussian interaction profile kernel similarity for miRNAs and diseases. This model constructed a heterogeneous graph consisting of three interlinked sub-graphs and further adopted depth-first search algorithm to infer potential miRNA-disease associations. As a result, PBMDA achieved reliable performance in the frameworks of both local and global LOOCV (AUCs of 0.8341 and 0.9169, respectively) and 5-fold cross validation (average AUC of 0.9172). In the cases studies of three important human diseases, 88% (Esophageal Neoplasms), 88% (Kidney Neoplasms) and 90% (Colon Neoplasms) of top-50 predicted miRNAs have been manually confirmed by previous experimental reports from literatures. Through the comparison performance between PBMDA and other previous models in case studies, the reliable performance also demonstrates that PBMDA could serve as a powerful computational tool to accelerate the identification of disease-miRNA associations. [ABSTRACT FROM AUTHOR]

Cancer researchers have long recognized that somatic mutations are not uniformly distributed within genes. However, most approaches for identifying cancer mutations focus on either the entire-gene or single amino-acid level. We have bridged these two methodologies with a multiscale mutation clustering algorithm that identifies variable length mutation clusters in cancer genes. We ran our algorithm on 539 genes using the combined mutation data in 23 cancer types from The Cancer Genome Atlas (TCGA) and identified 1295 mutation clusters. The resulting mutation clusters cover a wide range of scales and often overlap with many kinds of protein features including structured domains, phosphorylation sites, and known single nucleotide variants. We statistically associated these multiscale clusters with gene expression and drug response data to illuminate the functional and clinical consequences of mutations in our clusters. Interestingly, we find multiple clusters within individual genes that have differential functional associations: these include PTEN, FUBP1, and CDH1. This methodology has potential implications in identifying protein regions for drug targets, understanding the biological underpinnings of cancer, and personalizing cancer treatments. Toward this end, we have made the mutation clusters and the clustering algorithm available to the public. Clusters and pathway associations can be interactively browsed at . The multiscale mutation clustering algorithm is available at . [ABSTRACT FROM AUTHOR]

We present here a novel genetic algorithm-based random forest (GARF) modeling technique that enables a reduction in the complexity of large gene disease signatures to highly accurate, greatly simplified gene panels. When applied to 803 glioblastoma multiforme samples, this method allowed the 840-gene Verhaak et al. gene panel (the standard in the field) to be reduced to a 48-gene classifier, while retaining 90.91% classification accuracy, and outperforming the best available alternative methods. Additionally, using this approach we produced a 32-gene panel which allows for better consistency between RNA-seq and microarray-based classifications, improving cross-platform classification retention from 69.67% to 86.07%. A webpage producing these classifications is available at . [ABSTRACT FROM AUTHOR]

Nonnegative Matrix Factorization (NMF) has proved to be an effective method for unsupervised clustering analysis of gene expression data. By the nonnegativity constraint, NMF provides a decomposition of the data matrix into two matrices that have been used for clustering analysis. However, the decomposition is not unique. This allows different clustering results to be obtained, resulting in different interpretations of the decomposition. To alleviate this problem, some existing methods directly enforce uniqueness to some extent by adding regularization terms in the NMF objective function. Alternatively, various normalization methods have been applied to the factor matrices; however, the effects of the choice of normalization have not been carefully investigated. Here we investigate the performance of NMF for the task of cancer class discovery, under a wide range of normalization choices. After extensive evaluations, we observe that the maximum norm showed the best performance, although the maximum norm has not previously been used for NMF. Matlab codes are freely available from: . [ABSTRACT FROM AUTHOR]

Combining path consistency (PC) algorithms with conditional mutual information (CMI) are widely used in reconstruction of gene regulatory networks. CMI has many advantages over Pearson correlation coefficient in measuring non-linear dependence to infer gene regulatory networks. It can also discriminate the direct regulations from indirect ones. However, it is still a challenge to select the conditional genes in an optimal way, which affects the performance and computation complexity of the PC algorithm. In this study, we develop a novel conditional mutual information-based algorithm, namely RPNI (Regulation Pattern based Network Inference), to infer gene regulatory networks. For conditional gene selection, we define the co-regulation pattern, indirect-regulation pattern and mixture-regulation pattern as three candidate patterns to guide the selection of candidate genes. To demonstrate the potential of our algorithm, we apply it to gene expression data from DREAM challenge. Experimental results show that RPNI outperforms existing conditional mutual information-based methods in both accuracy and time complexity for different sizes of gene samples. Furthermore, the robustness of our algorithm is demonstrated by noisy interference analysis using different types of noise. [ABSTRACT FROM AUTHOR]

Four of the most common limitations of the many available clustering methods are: i) the lack of a proper strategy to deal with outliers; ii) the need for a good a priori estimate of the number of clusters to obtain reasonable results; iii) the lack of a method able to detect when partitioning of a specific data set is not appropriate; and iv) the dependence of the result on the initialization. Here we propose Cross-clustering (CC), a partial clustering algorithm that overcomes these four limitations by combining the principles of two well established hierarchical clustering algorithms: Ward’s minimum variance and Complete-linkage. We validated CC by comparing it with a number of existing clustering methods, including Ward’s and Complete-linkage. We show on both simulated and real datasets, that CC performs better than the other methods in terms of: the identification of the correct number of clusters, the identification of outliers, and the determination of real cluster memberships. We used CC to cluster samples in order to identify disease subtypes, and on gene profiles, in order to determine groups of genes with the same behavior. Results obtained on a non-biological dataset show that the method is general enough to be successfully used in such diverse applications. The algorithm has been implemented in the statistical language R and is freely available from the CRAN contributed packages repository. [ABSTRACT FROM AUTHOR]

DNA methylation is an important epigenetic event that effects gene expression during development and various diseases such as cancer. Understanding the mechanism of action of DNA methylation is important for downstream analysis. In the Illumina Infinium HumanMethylation 450K array, there are tens of probes associated with each gene. Given methylation intensities of all these probes, it is necessary to compute which of these probes are most representative of the gene centric methylation level. In this study, we developed a feature selection algorithm based on sequential forward selection that utilized different classification methods to compute gene centric DNA methylation using probe level DNA methylation data. We compared our algorithm to other feature selection algorithms such as support vector machines with recursive feature elimination, genetic algorithms and ReliefF. We evaluated all methods based on the predictive power of selected probes on their mRNA expression levels and found that a K-Nearest Neighbors classification using the sequential forward selection algorithm performed better than other algorithms based on all metrics. We also observed that transcriptional activities of certain genes were more sensitive to DNA methylation changes than transcriptional activities of other genes. Our algorithm was able to predict the expression of those genes with high accuracy using only DNA methylation data. Our results also showed that those DNA methylation-sensitive genes were enriched in Gene Ontology terms related to the regulation of various biological processes. [ABSTRACT FROM AUTHOR]

Reverse engineering approaches to constructing gene regulatory networks (GRNs) based on genome-wide mRNA expression data have led to significant biological findings, such as the discovery of novel drug targets. However, the reliability of the reconstructed GRNs needs to be improved. Here, we propose an ensemble-based network aggregation approach to improving the accuracy of network topologies constructed from mRNA expression data. To evaluate the performances of different approaches, we created dozens of simulated networks from combinations of gene-set sizes and sample sizes and also tested our methods on three Escherichia coli datasets. We demonstrate that the ensemble-based network aggregation approach can be used to effectively integrate GRNs constructed from different studies – producing more accurate networks. We also apply this approach to building a network from epithelial mesenchymal transition (EMT) signature microarray data and identify hub genes that might be potential drug targets. The R code used to perform all of the analyses is available in an R package entitled “ENA”, accessible on CRAN (). [ABSTRACT FROM AUTHOR]

Cancer and healthy cells have distinct distributions of molecular properties and thus respond differently to drugs. Cancer drugs ideally kill cancer cells while limiting harm to healthy cells. However, the inherent variance among cells in both cancer and healthy cell populations increases the difficulty of selective drug action. Here we formalize a classification framework based on the idea that an ideal cancer drug should maximally discriminate between cancer and healthy cells. More specifically, this discrimination should be performed on the basis of measurable cell markers. We divide the problem into three parts which we explore with examples. First, molecular markers should discriminate cancer cells from healthy cells at the single-cell level. Second, the effects of drugs should be statistically predicted by these molecular markers. Third, drugs should be optimized for classification performance. We find that expression levels of a handful of genes suffice to discriminate well between individual cells in cancer and healthy tissue. We also find that gene expression predicts the efficacy of some cancer drugs, suggesting that these cancer drugs act as suboptimal classifiers using gene profiles. Finally, we formulate a framework that defines an optimal drug, and predicts drug cocktails that may target cancer more accurately than the individual drugs alone. Conceptualizing cancer drugs as solving a discrimination problem in the high-dimensional space of molecular markers promises to inform the design of new cancer drugs and drug cocktails. [ABSTRACT FROM AUTHOR]

Transcriptome experiments are performed to assess protein abundance through mRNA expression analysis. Expression levels of genes vary depending on the experimental conditions and the cell response. Transcriptome data must be diverse and yet comparable in reference to stably expressed genes, even if they are generated from different experiments on the same biological context from various laboratories. In this study, expression patterns of 9090 microarray samples grouped into 381 NCBI-GEO datasets were investigated to identify novel candidate reference genes using randomizations and Receiver Operating Characteristic (ROC) curves. The analysis demonstrated that cell type specific reference gene sets display less variability than a united set for all tissues. Therefore, constitutively and stably expressed, origin specific novel reference gene sets were identified based on their coefficient of variation and percentage of occurrence in all GEO datasets, which were classified using Medical Subject Headings (MeSH). A large number of MeSH grouped reference gene lists are presented as novel tissue specific reference gene lists. The most commonly observed 17 genes in these sets were compared for their expression in 8 hepatocellular, 5 breast and 3 colon carcinoma cells by RT-qPCR to verify tissue specificity. Indeed, commonly used housekeeping genes GAPDH, Actin and EEF2 had tissue specific variations, whereas several ribosomal genes were among the most stably expressed genes in vitro. Our results confirm that two or more reference genes should be used in combination for differential expression analysis of large-scale data obtained from microarray or next generation sequencing studies. Therefore context dependent reference gene sets, as presented in this study, are required for normalization of expression data from diverse technological backgrounds. [ABSTRACT FROM AUTHOR]

The purpose of feature selection is to identify the relevant and non-redundant features from a dataset. In this article, the feature selection problem is organized as a graph-theoretic problem where a feature-dissimilarity graph is shaped from the data matrix. The nodes represent features and the edges represent their dissimilarity. Both nodes and edges are given weight according to the feature’s relevance and dissimilarity among the features, respectively. The problem of finding relevant and non-redundant features is then mapped into densest subgraph finding problem. We have proposed a multiobjective particle swarm optimization (PSO)-based algorithm that optimizes average node-weight and average edge-weight of the candidate subgraph simultaneously. The proposed algorithm is applied for identifying relevant and non-redundant disease-related genes from microarray gene expression data. The performance of the proposed method is compared with that of several other existing feature selection techniques on different real-life microarray gene expression datasets. [ABSTRACT FROM AUTHOR]

It has become increasingly clear that the current taxonomy of clinical phenotypes is mixed with molecular heterogeneity, which potentially affects the treatment effect for involved patients. Defining the hidden molecular-distinct diseases using modern large-scale genomic approaches is therefore useful for refining clinical practice and improving intervention strategies. Given that microRNA expression profiling has provided a powerful way to dissect hidden genetic heterogeneity for complex diseases, the aim of the study was to develop a bioinformatics approach that identifies microRNA features leading to the hidden subtyping of complex clinical phenotypes. The basic strategy of the proposed method was to identify optimal miRNA clusters by iteratively partitioning the sample and feature space using the two-ways super-paramagnetic clustering technique. We evaluated the obtained optimal miRNA cluster by determining the consistency of co-expression and the chromosome location among the within-cluster microRNAs, and concluded that the optimal miRNA cluster could lead to a natural partition of disease samples. We applied the proposed method to a publicly available microarray dataset of breast cancer patients that have notoriously heterogeneous phenotypes. We obtained a feature subset of 13 microRNAs that could classify the 71 breast cancer patients into five subtypes with significantly different five-year overall survival rates (45%, 82.4%, 70.6%, 100% and 60% respectively; p = 0.008). By building a multivariate Cox proportional-hazards prediction model for the feature subset, we identified has-miR-146b as one of the most significant predictor (p = 0.045; hazard ratios = 0.39). The proposed algorithm is a promising computational strategy for dissecting hidden genetic heterogeneity for complex diseases, and will be of value for improving cancer diagnosis and treatment. [ABSTRACT FROM AUTHOR]

Formalin fixed paraffin-embedded (FFPE) tumor specimens are the conventionally archived material in clinical practice, representing an invaluable tissue source for biomarkers development, validation and routine implementation. For many prospective clinical trials, this material has been collected allowing for a prospective-retrospective study design which represents a successful strategy to define clinical utility for candidate markers. Gene expression data can be obtained even from FFPE specimens with the broadly used Affymetrix HG-U133 Plus 2.0 microarray platform. Nevertheless, important major discrepancies remain in expression data obtained from FFPE compared to fresh-frozen samples, prompting the need for appropriate data processing which could help to obtain more consistent results in downstream analyses. In a publicly available dataset of matched frozen and FFPE expression data, the performances of different normalization methods and specifically designed Chip Description Files (CDFs) were compared. The use of an alternative CDFs together with fRMA normalization significantly improved frozen-FFPE sample correlations, frozen-FFPE probeset correlations and agreement of differential analysis between different tumor subtypes. The relevance of our optimized data processing was assessed and validated using two independent datasets. In this study we demonstrated that an appropriate data processing can significantly improve the reliability of gene expression data derived from FFPE tissues using the standard Affymetrix platform. Tools for the implementation of our data processing algorithm are made publicly available at http://www.biocut.unito.it/cdf-ffpe/. [ABSTRACT FROM AUTHOR]

We recently showed that differential expression of extracellular matrix (ECM) genes delineates four subgroups of breast carcinomas (ECM1, -2, -3- and -4) with different clinical outcome. To further investigate the characteristics of ECM signature and its impact on tumor progression, we conducted unsupervised clustering analyses in 6 additional independent datasets of invasive breast tumors from different platforms for a total of 643 samples. Use of four different clustering algorithms identified ECM3 tumors as an independent group in all datasets tested. ECM3 showed a homogeneous gene pattern, consisting of 58 genes encoding 43 structural ECM proteins. From 26 to 41% of the cases were ECM3-enriched, and analysis of datasets relevant to gene expression in neoplastic or corresponding stromal cells showed that both stromal and breast carcinoma cells can coordinately express ECM3 genes. In in vitro experiments, β-estradiol induced ECM3 gene production in ER-positive breast carcinoma cell lines, whereas TGFβ induced upregulation of the genes leading to ECM3 gene classification, especially in ER-negative breast carcinoma cells and in fibroblasts. Multivariate analysis of distant metastasis-free survival in untreated breast tumor patients revealed a significant interaction between ECM3 and histological grade (p = 0.001). Cox models, estimated separately in grade I–II and grade III tumors, indicated a highly significant association between ECM3 and worse survival probability only in grade III tumors (HR = 3.0, 95% CI = 1.3–7.0, p = 0.0098). Gene Set Enrichment analysis of ECM3 compared to non-ECM3 tumors revealed significant enrichment of epithelial-mesenchymal transition (EMT) genes in both grade I–II and grade III subsets of ECM3 tumors. Thus, ECM3 is a robust cluster that identifies breast carcinomas with EMT features but with accelerated metastatic potential only in the undifferentiated (grade III) phenotype. These findings support the key relevance of neoplastic and stroma interaction in breast cancer progression. [ABSTRACT FROM AUTHOR]

Identification and characterization of crucial gene target(s) that will allow focused therapeutics development remains a challenge. We have interrogated the putative therapeutic targets associated with the transcription factor Grainy head-like 2 (GRHL2), a critical epithelial regulatory factor. We demonstrate the possibility to define the molecular functions of critical genes in terms of their personalized expression profiles, allowing appropriate functional conclusions to be derived. A novel methodology, relative expression analysis with gene-set pairs (RXA-GSP), is designed to explore the potential clinical utility of cancer-biology discovery. Observing that Grhl2-overexpression leads to increased metastatic potential in vitro, we established a model assuming Grhl2-induced or -inhibited genes confer poor or favorable prognosis respectively for cancer metastasis. Training on public gene expression profiles of 995 breast cancer patients, this method prioritized one gene-set pair (GRHL2, CDH2, FN1, CITED2, MKI67 versus CTNNB1 and CTNNA3) from all 2717 possible gene-set pairs (GSPs). The identified GSP significantly dichotomized 295 independent patients for metastasis-free survival (log-rank tested p = 0.002; severe empirical p = 0.035). It also showed evidence of clinical prognostication in another independent 388 patients collected from three studies (log-rank tested p = 3.3e–6). This GSP is independent of most traditional prognostic indicators, and is only significantly associated with the histological grade of breast cancer (p = 0.0017), a GRHL2-associated clinical character (p = 6.8e–6, Spearman correlation), suggesting that this GSP is reflective of GRHL2-mediated events. Furthermore, a literature review indicates the therapeutic potential of the identified genes. This research demonstrates a novel strategy to integrate both biological experiments and clinical gene expression profiles for extracting and elucidating the genomic impact of a novel factor, GRHL2, and its associated gene-sets on the breast cancer prognosis. Importantly, the RXA-GSP method helps to individualize breast cancer treatment. It also has the potential to contribute considerably to basic biological investigation, clinical tools, and potential therapeutic targets. [ABSTRACT FROM AUTHOR]