With recent advances in microarray technology, there has been a flourish in genome-scale identification of molecular signatures for cancer. However, the differentially expressed genes obtained by different laboratories are highly divergent. The present discrepancy at gene level indicates a need for a novel strategy to obtain more robust signatures for cancer. In this paper we hypothesize that (1) the expression signatures of different cancer microarray datasets are more similar at pathway level than at gene level; (2) the comparability of the cancer molecular mechanisms of different individuals is related to their genetic similarities. In support of the hypotheses, we summarized theoretical and experimental evidences, and conducted case studies on colorectal and prostate cancer microarray datasets. Based on the above assumption, we propose that reliable cancer signatures should be investigated in the context of biological pathways, within a cohort of genetically homogeneous population. It is hoped that the hypotheses can guide future research in cancer mechanism and signature discovery.
Recent years have seen the development of various pathway-based methods for the analysis of microarray gene expression data. These approaches have the potential to bring biological insights into microarray studies. A variety of methods have been proposed to construct networks using gene expression data. Because individual pathways do not act in isolation, it is important to understand how different pathways coordinate to perform cellular functions. However, there are no published methods describing how to build pathway clusters that are closely related to traits of interest.
We propose to build pathway clusters from pathway-based classification methods. The proposed methods allow researchers to identify clusters of pathways sharing similar functions. These pathways may or may not share genes. As an illustration, our approach is applied to three human breast cancer microarray data sets. We found that our methods yielded consistent and interpretable results for these three data sets. We further investigated one of the pathway clusters found using PubMatrix. We found that informative genes in the pathway clusters do have more publications with keywords, like estrogen receptor, compared with informative genes in other top pathways. In addition, using the shortest path analysis in GeneGo's MetaCore and Human Protein Reference Database, we were able to identify the links which connect the pathways without shared genes within the pathway cluster.
Our proposed pathway clustering methods allow bioinformaticians and biologists to investigate how informative genes within pathways are related to each other and understand possible crosstalk between pathways in a cluster. Therefore, building pathway clusters may lead to a better understanding of molecular mechanisms affecting a trait of interest, and help generate further biological hypotheses from gene expression data.
Gene expression analysis has been intensively researched for more than a decade. Recently, there has been elevated interest in the integration of microarray data analysis with other types of biological knowledge in a holistic analytical approach. We propose a methodology that can be facilitated for pathway based microarray data analysis, based on the observation that a substantial proportion of genes present in biochemical pathway databases are members of a number of distinct pathways. Our methodology aims towards establishing the state of individual pathways, by identifying those truly affected by the experimental conditions based on the behaviour of such genes. For that purpose it considers all the pathways in which a gene participates and the general census of gene expression per pathway.
We utilise hill climbing, simulated annealing and a genetic algorithm to analyse the consistency of the produced results, through the application of fuzzy adjusted rand indexes and hamming distance. All algorithms produce highly consistent genes to pathways allocations, revealing the contribution of genes to pathway functionality, in agreement with current pathway state visualisation techniques, with the simulated annealing search proving slightly superior in terms of efficiency.
We show that the expression values of genes, which are members of a number of biochemical pathways or modules, are the net effect of the contribution of each gene to these biochemical processes. We show that by manipulating the pathway and module contribution of such genes to follow underlying trends we can interpret microarray results centred on the behaviour of these genes.
Gene expression technology has become a routine application in many laboratories and has provided large amounts of gene expression signatures that have been identified in a variety of cancer types. Interpretation of gene expression signatures would profit from the availability of a procedure capable of assigning differentially regulated genes or entire gene signatures to defined cancer signaling pathways. Here we describe a graph-based approach that identifies cancer signaling pathways from published gene expression signatures. Published gene expression signatures are collected in a database (PubLiME: Published Lists of Microarray Experiments) enabled for cross-platform gene annotation. Significant co-occurrence modules composed of up to 10 genes in different gene expression signatures are identified. Significantly co-occurring genes are linked by an edge in an undirected graph. Edge-betweenness and k-clique clustering combined with graph modularity as a quality measure are used to identify communities in the resulting graph. The identified communities consist of cell cycle, apoptosis, phosphorylation cascade, extra cellular matrix, interferon and immune response regulators as well as communities of unknown function. The genes constituting different communities are characterized by common genomic features and strongly enriched cis-regulatory modules in their upstream regulatory regions that are consistent with pathway assignment of those genes.
Expression profiling of whole genomes, and modern high-throughput proteomics, has created a revolution in the study of disease states. Approaches for gene expression analysis (time series analysis and clustering) have been applied to functional genomics related to cancer research, and have yielded major successes in the pursuit of gene expression signatures. However, these analysis methods are primarily designed to identify correlative or causal relationships between entities, but do not consider the data in the proper biological context of a “biological pathway” model. Pathway models form a cornerstone of systems biology. They provide a framework for (1) systematic interrogation of biochemical interactions, (2) management of the collective knowledge pertaining to cellular components, and (3) discovery of emergent properties of different pathway configurations.
CFD Research Corporation has developed advanced techniques to interpret microarray data in the context of known biological pathways. We have applied this integrative biological pathway-centered approach to the specific problem of identifying a genetic cause for individuals predisposed to mefloquine neurotoxicity. Mefloquine (Lariam) is highly effective against drug-resistant malaria. However, adverse neurological effects (ataxia, mood changes) have been observed in human sub-populations. Microarray experiments were used to quantify the transcriptional response of cells exposed to mefloquine. Canonical pathway models containing the differentially expressed genes were automatically retrieved from the KEGG database, using recently developed software. The canonical pathway models were automatically concatenated together to form the final pathway model. The resultant pathway model was interrogated using a novel signaling control flux (SCF) algorithm that combines Boolean pseudodynamics (BPD) to relax the cumbersome steady-state assumptions of SCF. The SCF-BPD algorithm was used to identify and prioritize pathways critical to adverse effects of mefloquine. Further analysis resulted in the identification of specific sub-cellular targets that may explain mefloquine neurotoxicity in human subpopulations on the basis of known single-nucleotide polymorphisms.
The analysis of microarray experiments requires accurate and up-to-date functional annotation of the microarray reporters to optimize the interpretation of the biological processes involved. Pathway visualization tools are used to connect gene expression data with existing biological pathways by using specific database identifiers that link reporters with elements in the pathways.
This paper proposes a novel method that aims to improve microarray reporter annotation by BLASTing the original reporter sequences against a species-specific EMBL subset, that was derived from and crosslinked back to the highly curated UniProt database. The resulting alignments were filtered using high quality alignment criteria and further compared with the outcome of a more traditional approach, where reporter sequences were BLASTed against EnsEMBL followed by locating the corresponding protein (UniProt) entry for the high quality hits. Combining the results of both methods resulted in successful annotation of > 58% of all reporter sequences with UniProt IDs on two commercial array platforms, increasing the amount of Incyte reporters that could be coupled to Gene Ontology terms from 32.7% to 58.3% and to a local GenMAPP pathway from 9.6% to 16.7%. For Agilent, 35.3% of the total reporters are now linked towards GO nodes and 7.1% on local pathways.
Our methods increased the annotation quality of microarray reporter sequences and allowed us to visualize more reporters using pathway visualization tools. Even in cases where the original reporter annotation showed the correct description the new identifiers often allowed improved pathway and Gene Ontology linking. These methods are freely available at http://www.bigcat.unimaas.nl/public/publications/Gaj_Annotation/.
Pathway or gene set analysis has become an increasingly popular approach for analyzing high-throughput biological experiments such as microarray gene expression studies. The purpose of pathway analysis is to identify differentially expressed pathways associated with outcomes. Important challenges in pathway analysis are selecting a subset of genes contributing most to association with clinical phenotypes and conducting statistical tests of association for the pathways efficiently. We propose a two-stage analysis strategy: (1) extract latent variables representing activities within each pathway using a dimension reduction approach based on adaptive elastic-net sparse principal component analysis; (2) integrate the latent variables with the regression modeling framework to analyze studies with different types of outcomes such as binary, continuous or survival outcomes. Our proposed approach is computationally efficient. For each pathway, because the latent variables are estimated in an unsupervised fashion without using disease outcome information, in the sample label permutation testing procedure, the latent variables only need to be calculated once rather than for each permutation resample. Using both simulated and real datasets, we show our approach performed favorably when compared with five other currently available pathway testing methods.
gene expression; microarray; pathway analysis; sparse principal component analysis
Numerous studies have used microarrays to identify gene signatures for predicting cancer patient clinical outcome and responses to chemotherapy. However, the potential impact of gene expression profiling in cancer diagnosis, prognosis and development of personalized treatment may not be fully exploited due to the lack of consensus gene signatures and poor understanding of the underlying molecular mechanisms.
We developed a novel approach to derive gene signatures for breast cancer prognosis in the context of known biological pathways. Using unsupervised methods, cancer patients were separated into distinct groups based on gene expression patterns in one of the following pathways: apoptosis, cell cycle, angiogenesis, metastasis, p53, DNA repair, and several receptor-mediated signaling pathways including chemokines, EGF, FGF, HIF, MAP kinase, JAK and NF-κB. The survival probabilities were then compared between the patient groups to determine if differential gene expression in a specific pathway is correlated with differential survival.
Our results revealed expression of cell cycle genes is strongly predictive of breast cancer outcomes. We further confirmed this observation by building a cell cycle gene signature model using supervised methods. Validated in multiple independent datasets, the cell cycle gene signature is a more accurate predictor for breast cancer clinical outcome than the previously identified Amsterdam 70-gene signature that has been developed into a FDA approved clinical test MammaPrint®.
Taken together, the gene expression signature model we developed from well defined pathways is not only a consistently powerful prognosticator but also mechanistically linked to cancer biology. Our approach provides an alternative to the current methodology of identifying gene expression markers for cancer prognosis and drug responses using the whole genome gene expression data.
The evaluation of statistical significance has become a critical process in identifying differentially expressed genes in microarray studies. Classical p-value adjustment methods for multiple comparisons such as family-wise error rate (FWER) have been found to be too conservative in analyzing large-screening microarray data, and the False Discovery Rate (FDR), the expected proportion of false positives among all positives, has been recently suggested as an alternative for controlling false positives. Several statistical approaches have been used to estimate and control FDR, but these may not provide reliable FDR estimation when applied to microarray data sets with a small number of replicates.
We propose a rank-invariant resampling (RIR) based approach to FDR evaluation. Our proposed method generates a biologically relevant null distribution, which maintains similar variability to observed microarray data. We compare the performance of our RIR-based FDR estimation with that of four other popular methods. Our approach outperforms the other methods both in simulated and real microarray data.
We found that the SAM's random shuffling and SPLOSH approaches were liberal and the other two theoretical methods were too conservative while our RIR approach provided more accurate FDR estimation than the other approaches.
A molecular characterization of Alzheimer's Disease (AD) is the key to the identification of altered gene sets that lead to AD progression. We rely on the assumption that candidate marker genes for a given disease belong to specific pathogenic pathways, and we aim at unveiling those pathways stable across tissues, treatments and measurement systems. In this context, we analyzed three heterogeneous datasets, two microarray gene expression sets and one protein abundance set, applying a recently proposed feature selection method based on regularization.
For each dataset we identified a signature that was successively evaluated both from the computational and functional characterization viewpoints, estimating the classification error and retrieving the most relevant biological knowledge from different repositories. Each signature includes genes already known to be related to AD and genes that are likely to be involved in the pathogenesis or in the disease progression. The integrated analysis revealed a meaningful overlap at the functional level.
The identification of three gene signatures showing a relevant overlap of pathways and ontologies, increases the likelihood of finding potential marker genes for AD.
Hypoxia is a condition of low oxygen tension occurring in the tumor and negatively correlated with the progression of the disease. We studied the gene expression profiles of nine neuroblastoma cell lines grown under hypoxic conditions to define gene signatures that characterize hypoxic neuroblastoma. The l1-l2
regularization applied to the entire transcriptome identified a single signature of 11 probesets discriminating the hypoxic state. We demonstrate that new hypoxia signatures, with similar discriminatory power, can be generated by a prior knowledge-based filtering in which a much smaller number of probesets, characterizing hypoxia-related biochemical pathways, are analyzed. l1-l2
regularization identified novel and robust hypoxia signatures within apoptosis, glycolysis, and oxidative phosphorylation Gene Ontology classes. We conclude that the filtering approach overcomes the noisy nature of the microarray data and allows generating robust signatures suitable for biomarker discovery and patients risk assessment in a fraction of computer time.
Motivation: There is great interest in pathway-based methods for genomics data analysis in the research community. Although machine learning methods, such as random forests, have been developed to correlate survival outcomes with a set of genes, no study has assessed the abilities of these methods in incorporating pathway information for analyzing microarray data. In general, genes that are identified without incorporating biological knowledge are more difficult to interpret. Correlating pathway-based gene expression with survival outcomes may lead to biologically more meaningful prognosis biomarkers. Thus, a comprehensive study on how these methods perform in a pathway-based setting is warranted.
Results: In this article, we describe a pathway-based method using random forests to correlate gene expression data with survival outcomes and introduce a novel bivariate node-splitting random survival forests. The proposed method allows researchers to identify important pathways for predicting patient prognosis and time to disease progression, and discover important genes within those pathways. We compared different implementations of random forests with different split criteria and found that bivariate node-splitting random survival forests with log-rank test is among the best. We also performed simulation studies that showed random forests outperforms several other machine learning algorithms and has comparable results with a newly developed component-wise Cox boosting model. Thus, pathway-based survival analysis using machine learning tools represents a promising approach in dissecting pathways and for generating new biological hypothesis from microarray studies.
Availability: R package Pwayrfsurvival is available from URL: http://www.duke.edu/∼hp44/pwayrfsurvival.htm
Supplementary information: Supplementary data are available at Bioinformatics online.
Classification of cancers based on gene expressions produces better accuracy
when compared to that of the clinical markers. Feature selection improves
the accuracy of these classification algorithms by reducing the chance
of overfitting that happens due to large number of features. We develop a
new feature selection method called Biological Pathway-based Feature Selection (BPFS) for microarray data. Unlike most of the existing methods,
our method integrates signaling and gene regulatory pathways with gene
expression data to minimize the chance of overfitting of the method and to
improve the test accuracy. Thus, BPFS selects a biologically meaningful feature
set that is minimally redundant. Our experiments on published breast
cancer datasets demonstrate that all of the top 20 genes found by our method
are associated with cancer. Furthermore, the classification accuracy of our
signature is up to 18% better than that of vant Veers 70 gene signature,
and it is up to 8% better accuracy than the best published feature selection
Gene set enrichment testing has helped bridge the gap from an individual gene to a systems biology interpretation of microarray data. Although gene sets are defined a priori based on biological knowledge, current methods for gene set enrichment testing treat all genes equal. It is well-known that some genes, such as those responsible for housekeeping functions, appear in many pathways, whereas other genes are more specialized and play a unique role in a single pathway. Drawing inspiration from the field of information retrieval, we have developed and present here an approach to incorporate gene appearance frequency (in KEGG pathways) into two current methods, Gene Set Enrichment Analysis (GSEA) and logistic regression-based LRpath framework, to generate more reproducible and biologically meaningful results.
Two breast cancer microarray datasets were analyzed to identify gene sets differentially expressed between histological grade 1 and 3 breast cancer. The correlation of Normalized Enrichment Scores (NES) between gene sets, generated by the original GSEA and GSEA with the appearance frequency of genes incorporated (GSEA-AF), was compared. GSEA-AF resulted in higher correlation between experiments and more overlapping top gene sets. Several cancer related gene sets achieved higher NES in GSEA-AF as well. The same datasets were also analyzed by LRpath and LRpath with the appearance frequency of genes incorporated (LRpath-AF). Two well-studied lung cancer datasets were also analyzed in the same manner to demonstrate the validity of the method, and similar results were obtained.
We introduce an alternative way to integrate KEGG PATHWAY information into gene set enrichment testing. The performance of GSEA and LRpath can be enhanced with the integration of appearance frequency of genes. We conclude that, generally, gene set analysis methods with the integration of information from KEGG PATHWAY performs better both statistically and biologically.
A number of publications have reported the use of microarray technology to identify gene expression signatures to infer mechanisms and pathways associated with systemic lupus erythematosus (SLE) in human peripheral blood mononuclear cells. However, meta-analysis approaches with microarray data have not been well-explored in SLE.
In this study, a pathway-based meta-analysis was applied to four independent gene expression oligonucleotide microarray data sets to identify gene expression signatures for SLE, and these data sets were confirmed by a fifth independent data set.
Differentially expressed genes (DEGs) were identified in each data set by comparing expression microarray data from control samples and SLE samples. Using Ingenuity Pathway Analysis software, pathways associated with the DEGs were identified in each of the four data sets. Using the leave one data set out pathway-based meta-analysis approach, a 37-gene metasignature was identified. This SLE metasignature clearly distinguished SLE patients from controls as observed by unsupervised learning methods. The final confirmation of the metasignature was achieved by applying the metasignature to a fifth independent data set.
The novel pathway-based meta-analysis approach proved to be a useful technique for grouping disparate microarray data sets. This technique allowed for validated conclusions to be drawn across four different data sets and confirmed by an independent fifth data set. The metasignature and pathways identified by using this approach may serve as a source for identifying therapeutic targets for SLE and may possibly be used for diagnostic and monitoring purposes. Moreover, the meta-analysis approach provides a simple, intuitive solution for combining disparate microarray data sets to identify a strong metasignature.
Please see Research Highlight: http://genomemedicine.com/content/3/5/30
Numerous studies used microarray gene expression data to extract metastasis-driving gene signatures for the prediction of breast cancer relapse. However, the accuracy and generality of the previously introduced biomarkers are not acceptable for reliable usage in independent datasets. This inadequacy is attributed to ignoring gene interactions by simple feature selection methods, due to their computational burden. In this study, an integrated approach with low computational cost was proposed for identifying a more predictive gene signature, for prediction of breast cancer recurrence. First, a small set of genes was primarily selected as signature by an appropriate filter feature selection (FFS) method. Then, a binary sub-class of protein-protein interaction (PPI) network was used to expand the primary set by adding adjacent proteins of each gene signature from the PPI-network. Subsequently, the support vector machine-based recursive feature elimination (SVMRFE) method was applied to the expression level of all the genes in the expanded set. Finally, the genes with the highest score by SVMRFE were selected as the new biomarkers. Accuracy of the final selected biomarkers was evaluated to classify four datasets on breast cancer patients, including 800 cases, into two cohorts of poor and good prognosis. The results of the five-fold cross validation test, using the support vector machine as a classifier, showed more than 13% improvement in the average accuracy, after modifying the primary selected signatures. Moreover, the method used in this study showed a lower computational cost compared to the other PPI-based methods. The proposed method demonstrated more robust and accurate biomarkers using the PPI network, at a low computational cost. This approach could be used as a supplementary procedure in microarray studies after applying various gene selection methods.
Breast cancer; feature selection method; protein–protein interaction; recurrence prediction; support vector machine
A statistically robust and biologically-based approach for analysis of microarray data is described that integrates independent biological knowledge and data with a global F-test for finding genes of interest that minimizes the need for replicates when used for hypothesis generation. First, each microarray is normalized to its noise level around zero. The microarray dataset is then globally adjusted by robust linear regression. Second, genes of interest that capture significant responses to experimental conditions are selected by finding those that express significantly higher variance than those expressing only technical variability. Clustering expression data and identifying expression-independent properties of genes of interest including upstream transcriptional regulatory elements (TREs), ontologies and networks or pathways organizes the data into a biologically meaningful system. We demonstrate that when the number of genes of interest is inconveniently large, identifying a subset of "beacon genes" representing the largest changes will identify pathways or networks altered by biological manipulation. The entire dataset is then used to complete the picture outlined by the "beacon genes." This allow construction of a structured model of a system that can generate biologically testable hypotheses. We illustrate this approach by comparing cells cultured on plastic or an extracellular matrix which organizes a dataset of over 2,000 genes of interest from a genome wide scan of transcription. The resulting model was confirmed by comparing the predicted pattern of TREs with experimental determination of active transcription factors.
Routine application of gene expression microarray technology is rapidly producing large amounts of data that necessitate new approaches of analysis. The analysis of a specific microarray experiment profits enormously from cross-comparing to other experiments. This process is generally performed by numerical meta-analysis of published data where the researcher chooses the datasets to be analyzed based on assumptions about the biological relations of published datasets to his own data, thus severely limiting the possibility of finding surprising connections. Here we propose using a repository of published gene lists for the identification of interesting datasets to be subjected to more detailed numerical analysis.
We have compiled lists of genes that have been reported as differentially regulated in cancer related microarray studies. We searched these gene lists for statistically significant overlaps with lists of genes regulated by the tumor suppressors p16 and pRB. We identified a highly significant overlap of p16 and pRB target genes with genes regulated by the EWS/FLI fusion protein. Detailed numerical analysis of these data identified two sets of genes with clearly distinct roles in the G1/S and the G2/M phases of the cell cycle, as measured by enrichment of Gene Ontology categories.
We show that mining of published gene lists in the absence of numerical detail about gene expression levels constitutes a fast, easy to perform, widely applicable, and unbiased route towards the identification of biologically related gene expression microarray datasets.
Publicly available data repositories facilitate the sharing of an ever-increasing amount of microarray data. However, these datasets remain highly underutilized. Reutilizing the data could offer insights into questions and diseases entirely distinct from those considered in the original experimental design.
We first analyzed microarray datasets derived from known perturbations of specific pathways using the samr package in R to identify specific patterns of change in gene expression. We refer to these pattern of gene expression alteration as a "pathway signatures." We then used Spearman's rank correlation coefficient, a non-parametric measure of correlation, to determine similarities between pathway signatures and disease profiles, and permutation analysis to evaluate false discovery rate. This enabled detection of statistically significant similarity between these pathway signatures and corresponding changes observed in human disease. Finally, we evaluated pathway activation, as indicated by correlation with the pathway signature, as a risk factor for poor prognosis using multiple unrelated, publicly available datasets.
We have developed a novel method, Expression-based Pathway Signature Analysis (EPSA). We demonstrate that ESPA is a rigorous computational approach for statistically evaluating the degree of similarity between highly disparate sources of microarray expression data. We also show how EPSA can be used in a number of cases to stratify patients with differential disease prognosis. EPSA can be applied to many different types of datasets in spite of different platforms, different experimental designs, and different species. Applying this method can yield new insights into human disease progression.
EPSA enables the use of publicly available data for an entirely new, translational purpose to enable the identification of potential pathways of dysregulation in human disease, as well as potential leads for therapeutic molecular targets.
Microarray technology is often used to identify the genes that are differentially expressed between two biological conditions. On the other hand, since microarray datasets contain a small number of samples and a large number of genes, it is usually desirable to identify small gene subsets with distinct pattern between sample classes. Such gene subsets are highly discriminative in phenotype classification because of their tightly coupling features. Unfortunately, such identified classifiers usually tend to have poor generalization properties on the test samples due to overfitting problem.
We propose a novel approach combining both supervised learning with unsupervised learning techniques to generate increasingly discriminative gene clusters in an iterative manner. Our experiments on both simulated and real datasets show that our method can produce a series of robust gene clusters with good classification performance compared with existing approaches.
This backward approach for refining a series of highly discriminative gene clusters for classification purpose proves to be very consistent and stable when applied to various types of training samples.
DNA microarray based gene expression analysis of primary breast cancers is not in general clinical use in spite of the widespread enthusiasm for the approach. However, several studies have identified gene expression patterns, or "signatures" that are predictive of disease outcome and assays based on these findings have become commercially available. Additional studies are required to further define gene signatures that can be used to help tailor the therapy best suited to for an individual patient. Recent gene expression profiling studies using cell line models to identify downstream transcriptional targets of oncogenic signaling pathways may help achieve this goal.
Identification of genes associated with cancer development and progression using microarray data is challenging because of the high dimensionality and cluster structure of gene expressions. Here the clusters are composed of multiple genes with coordinated biological functions and/or correlated expressions. In this article, we first propose a hybrid approach for clustering gene expressions. The hybrid approach uses both pathological pathway information and correlations of gene expressions. We propose using the group bridge, a novel clustering penalization approach, for analysis of cancer microarray data. The group bridge approach explicitly accounts for the cluster structure of gene expressions, and is capable of selecting gene clusters and genes within those selected clusters that are associated with cancer. We also develop an iterative algorithm for computing the group bridge estimator. Analysis of three cancer microarray datasets shows that the proposed approach can identify biologically meaningful gene clusters and genes within those identified clusters.
Gene selection; group bridge; microarray; clustering; penalization
The epidermal specific ablation of Trp53 gene leads to the spontaneous development of aggressive tumors in mice through a process that is accelerated by the simultaneous ablation of Rb gene. Since alterations of p53-dependent pathway are common hallmarks of aggressive, poor prognostic human cancers, these mouse models can recapitulate the molecular features of some of these human malignancies.
To evaluate this possibility, gene expression microarray analysis was performed in mouse samples. The mouse tumors display increased expression of cell cycle and chromosomal instability associated genes. Remarkably, they are also enriched in human embryonic stem cell gene signatures, a characteristic feature of human aggressive tumors. Using cross-species comparison and meta-analytical approaches, we also observed that spontaneous mouse tumors display robust similarities with gene expression profiles of human tumors bearing mutated TP53, or displaying poor prognostic outcome, from multiple body tissues. We have obtained a 20-gene signature whose genes are overexpressed in mouse tumors and can identify human tumors with poor outcome from breast cancer, astrocytoma and multiple myeloma. This signature was consistently overexpressed in additional mouse tumors using microarray analysis. Two of the genes of this signature, AURKA and UBE2C, were validated in human breast and cervical cancer as potential biomarkers of malignancy.
Our analyses demonstrate that these mouse models are promising preclinical tools aimed to search for malignancy biomarkers and to test targeted therapies of prospective use in human aggressive tumors and/or with p53 mutation or inactivation.
Gene set enrichment analysis (GSEA) is an analytic approach which simultaneously reduces the dimensionality of microarray data and enables ready inference of the biological meaning of observed gene expression patterns. Here we invert the GSEA process to identify class-specific gene signatures. Because our approach uses the Kolmogorov-Smirnov approach both to define class specific signatures and to classify samples using those signatures, we have termed this methodology “Dual-KS” (DKS).
The optimum gene signature identified by the DKS algorithm was smaller than other methods to which it was compared in 5 out of 10 datasets. The estimated error rate of DKS using the optimum gene signature was smaller than the estimated error rate of the random forest method in 4 out of the 10 datasets, and was equivalent in two additional datasets. DKS performance relative to other benchmarked algorithms was similar to its performance relative to random forests.
DKS is an efficient analytic methodology that can identify highly parsimonious gene signatures useful for classification in the context of microarray studies. The algorithm is available as the dualKS package for R as part of the bioconductor project.
gene set enrichment analysis (GSEA); gene expression; DKS algorithm; gene signatures
Microarray experiments are becoming increasingly common in biomedical research, as is their deposition in publicly accessible repositories, such as Gene Expression Omnibus (GEO). As such, there has been a surge in interest to use this microarray data for meta-analytic approaches, whether to increase sample size for a more powerful analysis of a specific disease (e.g. lung cancer) or to re-examine experiments for reasons different than those examined in the initial, publishing study that generated them. For the average biomedical researcher, there are a number of practical barriers to conducting such meta-analyses such as manually aggregating, filtering and formatting the data. Methods to automatically process large repositories of microarray data into a standardized, directly comparable format will enable easier and more reliable access to microarray data to conduct meta-analyses.
We present a straightforward, simple but robust against potential outliers method for automatic quality control and pre-processing of tens of thousands of single-channel microarray data files. GEO GDS files are quality checked by comparing parametric distributions and quantile normalized to enable direct comparison of expression level for subsequent meta-analyses.
13,000 human 1-color experiments were processed to create a single gene expression matrix that subsets can be extracted from to conduct meta-analyses. Interestingly, we found that when conducting a global meta-analysis of gene-gene co-expression patterns across all 13,000 experiments to predict gene function, normalization had minimal improvement over using the raw data.
Normalization of microarray data appears to be of minimal importance on analyses based on co-expression patterns when the sample size is on the order of thousands microarray datasets. Smaller subsets, however, are more prone to aberrations and artefacts, and effective means of automating normalization procedures not only empowers meta-analytic approaches, but aids in reproducibility by providing a standard way of approaching the problem.
Data availability: matrix containing normalized expression of 20,813 genes across 13,000 experiments is available for download at . Source code for GDS files pre-processing is available from the authors upon request.