Gene Expression Classification Research Articles

A new approach to enhance the performance of decision tree for classifying gene expression data

BackgroundGene expression data classification is a challenging task due to the large dimensionality and very small number of samples. Decision tree is one of the popular machine learning approaches to address such classification problems. However, the existing decision tree algorithms use a single gene feature at each node to split the data into its child nodes and hence might suffer from poor performance specially when classifying gene expression dataset.ResultsBy using a new decision tree algorithm where, each node of the tree consists of more than one gene, we enhance the classification performance of traditional decision tree classifiers. Our method selects suitable genes that are combined using a linear function to form a derived composite feature. To determine the structure of the tree we use the area under the Receiver Operating Characteristics curve (AUC). Experimental analysis demonstrates higher classification accuracy using the new decision tree compared to the other existing decision trees in literature.ConclusionWe experimentally compare the effect of our scheme against other well known decision tree techniques. Experiments show that our algorithm can substantially boost the classification performance of the decision tree.

Michigan vs. Pittsburgh Style GA Optimisation of Fuzzy Rule Bases for Gene Expression Analysis

Microarray studies and gene expression analysis have received a lot of attention and provide many promising avenues towards the understanding of fundamental questions in biology and medicine. In this paper, the authors perform gene expression analysis and apply two hybrid GA-fuzzy approaches to classify gene expression data. Both are based on fuzzy if-then rule bases but they differ in the way these rule bases are optimised. The authors employ both a Michigan style approach, where single rules are handled as individuals in the population of the genetic algorithm, and a Pittsburgh type algorithm, which treats whole rule sets as individuals. Experimental results show that both approaches achieve good classification accuracy but that the Michigan style algorithm clearly outperforms the Pittsburgh classifier.

Unbiased transcriptome signature of in vivo cell proliferation reveals pro- and antiproliferative gene networks

Different types of mature B-cell lymphocytes are overall highly similar. Nevertheless, some B cells proliferate intensively, while others rarely do. Here, we demonstrate that a simple binary classification of gene expression in proliferating vs. resting B cells can identify, with remarkable selectivity, global in vivo regulators of the mammalian cell cycle, many of which are also post-translationally regulated by the APC/C E3 ligase. Consequently, we discover a novel regulatory network between the APC/C and the E2F transcription factors and discuss its potential impact on the G1–S transition of the cell cycle. In addition, by focusing on genes whose expression inversely correlates with proliferation, we demonstrate the inherent ability of our approach to also identify in vivo regulators of cell differentiation, cell survival, and other antiproliferative processes. Relying on data sets of wt, non-transgenic animals, our approach can be applied to other cell lineages and human data sets.

Sparse Manifold Clustering and Embedding to discriminate gene expression profiles of glioblastoma and meningioma tumors

Sparse Manifold Clustering and Embedding (SMCE) algorithm has been recently proposed for simultaneous clustering and dimensionality reduction of data on nonlinear manifolds using sparse representation techniques. In this work, SMCE algorithm is applied to the differential discrimination of Glioblastoma and Meningioma Tumors by means of their Gene Expression Profiles. Our purpose was to evaluate the robustness of this nonlinear manifold to classify gene expression profiles, characterized by the high-dimensionality of their representations and the low discrimination power of most of the genes. For this objective, we used SMCE to reduce the dimensionality of a preprocessed dataset of 35 single-labeling cDNA microarrays with 11500 original clones. Afterwards, supervised and unsupervised methodologies were applied to obtain the classification model: the former was based on linear discriminant analysis, the later on clustering using the SMCE embedding data. The results obtained using both approaches showed that all (100%) the samples could be correctly classified and the results of all repetitions but one formed a compatible cluster of predictive labels. Finally, the embedding dimensionality of the dataset extracted by SMCE revealed large discrimination margins between both classes.

A simulation to analyze feature selection methods utilizing gene ontology for gene expression classification

Gene expression profile classification is a pivotal research domain assisting in the transformation from traditional to personalized medicine. A major challenge associated with gene expression data classification is the small number of samples relative to the large number of genes. To address this problem, researchers have devised various feature selection algorithms to reduce the number of genes. Recent studies have been experimenting with the use of semantic similarity between genes in Gene Ontology (GO) as a method to improve feature selection. While there are few studies that discuss how to use GO for feature selection, there is no simulation study that addresses when to use GO-based feature selection. To investigate this, we developed a novel simulation, which generates binary class datasets, where the differentially expressed genes between two classes have some underlying relationship in GO. This allows us to investigate the effects of various factors such as the relative connectedness of the underlying genes in GO, the mean magnitude of separation between differentially expressed genes denoted by δ, and the number of training samples. Our simulation results suggest that the connectedness in GO of the differentially expressed genes for a biological condition is the primary factor for determining the efficacy of GO-based feature selection. In particular, as the connectedness of differentially expressed genes increases, the classification accuracy improvement increases. To quantify this notion of connectedness, we defined a measure called Biological Condition Annotation Level BCAL(G), where G is a graph of differentially expressed genes. Our main conclusions with respect to GO-based feature selection are the following: (1) it increases classification accuracy when BCAL(G)⩾0.696; (2) it decreases classification accuracy when BCAL(G)⩽0.389; (3) it provides marginal accuracy improvement when 0.389<BCAL(G)<0.696 and δ<1; (4) as the number of genes in a biological condition increases beyond 50 and δ⩾0.7, the improvement from GO-based feature selection decreases; and (5) we recommend not using GO-based feature selection when a biological condition has less than ten genes. Our results are derived from datasets preprocessed using RMA (Robust Multi-array Average), cases where δ is between 0.3 and 2.5, and training sample sizes between 20 and 200, therefore our conclusions are limited to these specifications. Overall, this simulation is innovative and addresses the question of when SoFoCles-style feature selection should be used for classification instead of statistical-based ranking measures.

Genome‐wide transcript analysis of early maize leaf development reveals gene cohorts associated with the differentiation of C4 Kranz anatomy

Photosynthesis underpins the viability of most ecosystems, with C4 plants that exhibit 'Kranz' anatomy being the most efficient primary producers. Kranz anatomy is characterized by closely spaced veins that are encircled by two morphologically distinct photosynthetic cell types. Although Kranz anatomy evolved multiple times, the underlying genetic mechanisms remain largely elusive, with only the maize scarecrow gene so far implicated in Kranz patterning. To provide a broader insight into the regulation of Kranz differentiation, we performed a genome-wide comparative analysis of developmental trajectories in Kranz (foliar leaf blade) and non-Kranz (husk leaf sheath) leaves of the C4 plant maize. Using profile classification of gene expression in early leaf primordia, we identified cohorts of genes associated with procambium initiation and vascular patterning. In addition, we used supervised classification criteria inferred from anatomical and developmental analyses of five developmental stages to identify candidate regulators of cell-type specification. Our analysis supports the suggestion that Kranz anatomy is patterned, at least in part, by a SCARECROW/SHORTROOT regulatory network, and suggests likely components of that network. Furthermore, the data imply a role for additional pathways in the development of Kranz leaves.

Incorporating feature ranking and evolutionary methods for the classification of high-dimensional DNA microarray gene expression data

DNA microarray gene expression classification poses a challenging task to the machine learning domain. Typically, the dimensionality of gene expression data sets could go from several thousands to over 10,000 genes. A potential solution to this issue is using feature selection to reduce the dimensionality. The aim of this paper is to investigate how we can use feature quality information to improve the precision of microarray gene expression classification tasks. We propose two evolutionary machine learning models based on the eXtended Classifier System (XCS) and a typical feature selection methodology. The first one, which we call FS-XCS, uses feature selection for feature reduction purposes. The second model is GRD-XCS, which uses feature ranking to bias the rule discovery process of XCS. The results indicate that the use of feature selection/ranking methods is essential for tackling highdimensional classification tasks, such as microarray gene expression classification. However, the results also suggest that using feature ranking to bias the rule discovery process performs significantly better than using the feature reduction method. In other words, using feature quality information to develop a smarter learning procedure is more efficient than reducing the feature set. Our findings have shown that extracting feature quality information can assist the learning process and improve classification accuracy. On the other hand, relying exclusively on the feature quality information might potentially decrease the classification performance (e.g., using feature reduction). Therefore, we recommend a hybrid approach that uses feature quality information to direct the learning process by highlighting the more informative features, but at the same time not restricting the learning process to explore other features.

Gene Expression Classification of Colon Cancer into Molecular Subtypes: Characterization, Validation, and Prognostic Value

BackgroundColon cancer (CC) pathological staging fails to accurately predict recurrence, and to date, no gene expression signature has proven reliable for prognosis stratification in clinical practice, perhaps because CC is a heterogeneous disease. The aim of this study was to establish a comprehensive molecular classification of CC based on mRNA expression profile analyses.Methods and FindingsFresh-frozen primary tumor samples from a large multicenter cohort of 750 patients with stage I to IV CC who underwent surgery between 1987 and 2007 in seven centers were characterized for common DNA alterations, including BRAF, KRAS, and TP53 mutations, CpG island methylator phenotype, mismatch repair status, and chromosomal instability status, and were screened with whole genome and transcriptome arrays. 566 samples fulfilled RNA quality requirements. Unsupervised consensus hierarchical clustering applied to gene expression data from a discovery subset of 443 CC samples identified six molecular subtypes. These subtypes were associated with distinct clinicopathological characteristics, molecular alterations, specific enrichments of supervised gene expression signatures (stem cell phenotype–like, normal-like, serrated CC phenotype–like), and deregulated signaling pathways. Based on their main biological characteristics, we distinguished a deficient mismatch repair subtype, a KRAS mutant subtype, a cancer stem cell subtype, and three chromosomal instability subtypes, including one associated with down-regulated immune pathways, one with up-regulation of the Wnt pathway, and one displaying a normal-like gene expression profile. The classification was validated in the remaining 123 samples plus an independent set of 1,058 CC samples, including eight public datasets. Furthermore, prognosis was analyzed in the subset of stage II–III CC samples. The subtypes C4 and C6, but not the subtypes C1, C2, C3, and C5, were independently associated with shorter relapse-free survival, even after adjusting for age, sex, stage, and the emerging prognostic classifier Oncotype DX Colon Cancer Assay recurrence score (hazard ratio 1.5, 95% CI 1.1–2.1, p = 0.0097). However, a limitation of this study is that information on tumor grade and number of nodes examined was not available.ConclusionsWe describe the first, to our knowledge, robust transcriptome-based classification of CC that improves the current disease stratification based on clinicopathological variables and common DNA markers. The biological relevance of these subtypes is illustrated by significant differences in prognosis. This analysis provides possibilities for improving prognostic models and therapeutic strategies. In conclusion, we report a new classification of CC into six molecular subtypes that arise through distinct biological pathways. Please see later in the article for the Editors' Summary

Abstract 806: RNA sequencing based prognostic indexes show a high degree of correlation with historic array based gene expression measurements in multiple myeloma.

Abstract The Multiple Myeloma Research Foundation (MMRF) and the Multiple Myeloma Research Consortium (MMRC) have pioneered efforts towards understanding the genomic underpinnings of multiple myeloma. The MMRC centers have submitted over 1900 tumor samples to the MMRC tissue bank, from which &amp;gt;400 have been used in the Multiple Myeloma Genomics Initiative (MMGI), which is a multi-institutional comprehensive study to unveil the complete genomic blueprint of myeloma with public data access. In the initial phase 238 myeloma tumors were analyzed for somatic copy number alterations using array-based comparative genomic hybridization (aCGH) and array-based gene expression profiling (GEP). The large majority of this reference collection was also analyzed by array-based methylation profiling. We have also performed whole genome or exome sequencing on &amp;gt;200 normal/tumor pairs. These studies led to several key discoveries including global hypomethylation, homozygous deletion of UTX, novel mutations in FAM46C (13%) and DIS3/RRP44 (11%), and potentially therapeutically actionable BRAF mutations (4%). To improve our understanding of the transcriptional activities in myeloma, we have performed RNA-sequencing on 84 myeloma patient samples and 65 human myeloma cell lines for which whole genome or exome sequencing data is also available. RNA-seq has a number of advantages over GEP including improved dynamic range and the detection of exon and isoform levels, allele specific expression, and fusion transcripts. There are a number of robust gene expression classifications and indexes with prognostic value for myeloma samples tested by GEP. Within the cohort of 84 patient samples there are 42 samples with matched GEP data. Using this subgroup we tested if the current GEP methods could be directly moved to RNA-seq data by matching the current probesets to genes. We then compared the results across the patient set for each index or classification. For classifications we determined that the TC classification could be directly transitioned to RNA-seq data with 100% concordance. For the indexes we showed a strong correlation for the proliferation index (R2=0.971) and the NFKB index (R2=0.961) but only a moderate correlation for the 70-gene index (R2=0.761). The decreased correlation in the 70-gene index is clearly due to the large number of probesets used, which are associated with genes that are clearly not expressed by RNA-seq. Integration of the DNA and RNA sequencing data showed that 42% of the mutations detected in the DNA are expressed. Interestingly we observed the progressive loss of wildtype FAM46C alleles, which are often heterozygously mutated and expressed in patients but is exclusively homozogous in cell lines. Additional integrated analysis will improve out understanding of myelomagenesis and might lead to new and improved classification and prognostic models. Citation Format: Jonathan Keats, Venkata Yellapantula, Winnie Liang, Ahmet Kurdoglu, Angela Baker, Joan Levy, Todd Golub, David Craig, John Carpten. RNA sequencing based prognostic indexes show a high degree of correlation with historic array based gene expression measurements in multiple myeloma. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 806. doi:10.1158/1538-7445.AM2013-806

Classification of Time Series Gene Expression in Clinical Studies via Integration of Biological Network

The increasing availability of time series expression datasets, although promising, raises a number of new computational challenges. Accordingly, the development of suitable classification methods to make reliable and sound predictions is becoming a pressing issue. We propose, here, a new method to classify time series gene expression via integration of biological networks. We evaluated our approach on 2 different datasets and showed that the use of a hidden Markov model/Gaussian mixture models hybrid explores the time-dependence of the expression data, thereby leading to better prediction results. We demonstrated that the biclustering procedure identifies function-related genes as a whole, giving rise to high accordance in prognosis prediction across independent time series datasets. In addition, we showed that integration of biological networks into our method significantly improves prediction performance. Moreover, we compared our approach with several state-of–the-art algorithms and found that our method outperformed previous approaches with regard to various criteria. Finally, our approach achieved better prediction results on early-stage data, implying the potential of our method for practical prediction.

Increased expression of c-jun in nonalcoholic fatty liver disease

The lack of animal models which closely mimic the human situation is one main reason why the molecular mechanisms causing progression from nonalcoholic fatty liver to (NASH) still are elusive. In addition to the quantity of consumed calories, certain dietary components, namely saturated fatty acids, fructose, and cholesterol are increasingly recognized as dietary factors affecting NAFLD progression. The aim of this study was to developed and characterize a NASH inducing diet (ND) containing these nutrients in a composition which mimics Western food. Methods and Results: Feeding ND to male C57Bl6/N mice for 12 weeks induced significant body weight gain, an impaired glucose tolerance and elevated fasting blood glucose levels as well as decreased adiponectin and increased leptin serum levels. Furthermore, ND-feeding led to marked hepatic steatosis and distinct signs of hepatic oxidative stress and lipid peroxidation. In line with this, ND-fed mice exhibited more apoptotic hepatocytes and elevated transaminase serum levels as well as hepatic inflammation, activation of hepatic stellate cells, and beginning fibrosis, as assessed by histology and hydroxyprolin analysis. Transcriptome-wide gene expression analysis and classification in Gene Ontology categories revealed main alterations in the categories „lipid metabolic process”, “innate immune response” and “inflammatory response“ in the liver of HD-fed mice, which closely resembled NASH patients. Search for over-represented transcription factor target sites among the differenzially expressed genes identified AP-1 as the most likely factor to cause the transcriptional changes. Interestingly, protein-protein interaction network analysis identified c-JUN (a component if the AP-1 complex) as hub. Westernblot and immunohistochemical analysis confirmed c-jun-phosphorylation and nuclear translocation in murine ND-livers as well as NASH patients. Unexpectedly, also expression of (unphosphorylated) c-jun was significantly enhanced not only in NASH, but also in fatty liver without inflammation. Immunohistochemical analysis of a tissue micro array containing 80 human NAFLD tissues revealed a significant correlation between c-jun expression and the extent of hepatic steatosis. Conclusions: Our novel murine NASH-model induces important pathomechanisms also found in human NASH and led to the identification of increased c-jun already in simple steatosis. It is likely that higher abundance of this crucial pro-inflammatory regulator facilitates NASH upon additional stimuli. The identification of the latter as well as c-jun itself appear as attractive prognostic and therapeutic targets of NAFLD progression.

A Comparative Study of Feature Selection and Classification Methods for Gene Expression Data of Glioma

Microarray gene expression data gained great importance in recent years due to its role in disease diagnoses and prognoses which help to choose the appropriate treatment plan for patients. This technology has shifted a new era in molecular classification. Interpreting gene expression data remains a difficult problem and an active research area due to their native nature of “high dimensional low sample size”. Such problems pose great challenges to existing classification methods. Thus, effective feature selection techniques are often needed in this case to aid to correctly classify different tumor types and consequently lead to a better understanding of genetic signatures as well as improve treatment strategies. This paper aims on a comparative study of state-of-the- art feature selection methods, classification methods, and the combination of them, based on gene expression data. We compared the efficiency of three different classification methods including: support vector machines, k-nearest neighbor and random forest, and eight different feature selection methods, including: information gain, twoing rule, sum minority, max minority, gini index, sum of variances, t-statistics, and one-dimension support vector machine. Five-fold cross validation was used to evaluate the classification performance. Two publicly available gene expression data sets of glioma were used in the experiments. Results revealed the important role of feature selection in classifying gene expression data. By performing feature selection, the classification accuracy can be significantly boosted by using a small number of genes. The relationship of features selected in different feature selection methods is investigated and the most frequent features selected in each fold among all methods for both datasets are evaluated.

Empirical Evidence of the Applicability of Functional Clustering through Gene Expression Classification

The availability of a great range of prior biological knowledge about the roles and functions of genes and gene-gene interactions allows us to simplify the analysis of gene expression data to make it more robust, compact, and interpretable. Here, we objectively analyze the applicability of functional clustering for the identification of groups of functionally related genes. The analysis is performed in terms of gene expression classification and uses predictive accuracy as an unbiased performance measure. Features of biological samples that originally corresponded to genes are replaced by features that correspond to the centroids of the gene clusters and are then used for classifier learning. Using 10 benchmark data sets, we demonstrate that functional clustering significantly outperforms random clustering without biological relevance. We also show that functional clustering performs comparably to gene expression clustering, which groups genes according to the similarity of their expression profiles. Finally, the suitability of functional clustering as a feature extraction technique is evaluated and discussed.

Microarray gene expression classification with few genes: Criteria to combine attribute selection and classification methods

Microarray data classification is a task involving high dimensionality and small samples sizes. A common criterion to decide on the number of selected genes is maximizing the accuracy, which risks overfitting and usually selects more genes than actually needed. We propose, relaxing the maximum accuracy criterion, to select the combination of attribute selection and classification algorithm that using less attributes has an accuracy not statistically significantly worst that the best. Also we give some advice to choose a suitable combination of attribute selection and classifying algorithms for a good accuracy when using a low number of gene expressions. We used some well known attribute selection methods (FCBF, ReliefF and SVM-RFE, plus a Random selection, used as a base line technique) and classifying techniques (Naive Bayes, 3 Nearest Neighbor and SVM with linear kernel) applied to 30 data sets involving different cancer types.

Gene expression classification using binary rule majority voting genetic programming classifier

The results of a gene expression study are difficult to interpret. To increase interpretability, researchers have developed classification techniques that produce rules to classify gene expression profiles. Genetic programming is one method to produce classification rules. These rules are difficult to interpret because they are based on complicated functions of gene expression values. We propose the binary rule majority voting genetic programming classifier BRMVGPC that classifies samples using binary rules based on the detection calls for genes instead of the gene expression values. BRMVGPC increases rule interpretability. We evaluate BRMVGPC on two public datasets, one brain and one prostate cancer, and achieved 88.89% and 86.39% accuracy respectively. These results are comparable to other classifiers in the gene expression profile domain. Specific contributions include a classification technique BRMVGPC and an iterative k-nearest neighbour technique for handling marginal detection call values.

TNBCtype: A Subtyping Tool for Triple-Negative Breast Cancer

Motivation:Triple-negative breast cancer (TNBC) is a heterogeneous breast cancer group, and identification of molecular subtypes is essential for understanding the biological characteristics and clinical behaviors of TNBC as well as for developing personalized treatments. Based on 3,247 gene expression profiles from 21 breast cancer data sets, we discovered six TNBC subtypes from 587 TNBC samples with unique gene expression patterns and ontologies. Cell line models representing each of the TNBC subtypes also displayed different sensitivities to targeted therapeutic agents. Classification of TNBC into subtypes will advance further genomic research and clinical applications.Result:We developed a web-based subtyping tool TNBCtype for candidate TNBC samples using our gene expression meta data and classification methods. Given a gene expression data matrix, this tool will display for each candidate sample the predicted subtype, the corresponding correlation coefficient, and the permutation P-value. We offer a user-friendly web interface to predict the subtypes for new TNBC samples that may facilitate diagnostics, biomarker selection, drug discovery, and the more tailored treatment of breast cancer.

A fuzzy intelligent approach to the classification problem in gene expression data analysis

Classification is an important data mining task that widely used in several different real world applications. In microarray analysis, classification techniques are applied in order to discriminate diseases or to predict outcomes based on gene expression patterns, and perhaps even to identify the best treatment for given genetic signature. The most important challenge in gene expression data analysis lies in how to deal with its unique “high dimension small sample” characteristic, which makes many traditional classification techniques non-applicable or inefficient; and hence, more dedicated techniques are nowadays needed in order to approach this problem. Fuzzy logic is recently shown that is a powerful and suitable soft computing tool for handling the complex problems under incomplete data conditions. In this paper, a new hybrid model is proposed that combines artificial intelligence with fuzzy in order to benefit from unique advantages of both fuzzy logic and the classification power of the artificial neural networks (ANNs), to construct an efficient and accurate hybrid classifier in less available data situations. The proposed model, because of using the fuzzy parameters instead of the crisp parameters, will need less data set in comparing with traditional nonfuzzy neural networks in its training process or with same training sample can better learn and hence can yield more accurate results than traditional neural networks. In addition of theoretical evidence of using fuzzy logic, empirical results of gene expression classification indicate that the proposed model exhibits effectively improved classification accuracy in comparison with traditional artificial neural networks (ANNs) and also some other well-known statistical and intelligent classification models such as the linear discriminant analysis (LDA), the quadratic discriminant analysis (QDA), the K-nearest neighbor (KNN), and the support vector machines (SVMs). Therefore, the proposed model can be applied as an appropriate alternate approach for solving problems with scant data such as gene expression data classification, specifically when higher classification accuracy is needed.

Integrating pathway analysis and genetics of gene expression for genome-wide association study of basal cell carcinoma.

Genome-wide association studies (GWASs) have primarily focused on marginal effects for individual markers and have incorporated external functional information only after identifying robust statistical associations. We applied a new approach combining the genetics of gene expression and functional classification of genes to the GWAS of basal cell carcinoma (BCC) to identify potential biological pathways associated with BCC. We first identified 322,324 expression-associated single-nucleotide polymorphisms (eSNPs) from two existing GWASs of global gene expression in lymphoblastoid cell lines (n=955), and evaluated the association of these functionally annotated SNPs with BCC among 2,045 BCC cases and 6,013 controls in Caucasians. We then grouped them into 99 KEGG pathways for pathway analysis and identified two pathways associated with BCC with p value <0.05 and false discovery rate (FDR) <0.5: the autoimmune thyroid disease pathway (mainly HLA class I and II antigens, p<0.001, FDR=0.24) and Janus kinase-signal transducer and activator of transcription (JAK-STAT) signaling pathway (p=0.02, FDR=0.49). Seventy-nine (25.7%) out of 307 significant eSNPs in the JAK-STAT pathway were associated with BCC risk (p<0.05) in an independent replication set of 278 BCC cases and 1,262 controls. In addition, the association of JAK-STAT signaling pathway was marginally validated using 16,691 eSNPs identified from 110 normal skin samples (p=0.08). Based on the evidence of biological functions of the JAK-STAT pathway on oncogenesis, it is plausible that this pathway is involved in BCC pathogenesis.

Abstract A42: Autocrine-produced IL-6 and IL-8 are essential for the growth and maintenance of the transformed phenotype of basal-like breast cancer cells

Abstract Breast cancers are now subdivided into distinct tumor types based upon molecular classification of tumor gene expression. While most breast cancers are driven by estrogen-mediated signaling (estrogen receptor-positive, (ER+) or Luminal-type breast cancers) or by addiction to the HER2 oncogene (HER2+ breast cancer), the essential cellular signals that control the growth of ER-negative and HER2-negative (or basal-type) breast cancers (approximately 15% of all breast cancers), are incompletely understood. Patients with these ER− and HER2− breast cancers have poor outcomes and few available treatment options other than non-specific chemotherapy. To identify novel targets for the treatment and prevention of basal breast cancers, we investigated the activity of inflammatory genes in these cancers, previously linked to the progression of several other types of cancer. To identify critical inflammatory genes in basal breast cancer, we performed univariate t-tests (FDR&amp;lt;.05) on RNA profiles from the Baylor College of Medicine tumor bank (N=106) to identify inflammatory genes (from 226 genes in the KEGG Cytokine-Cytokine Receptor set) that are differentially expressed in ER-negative breast cancers as compared to ER+ or luminal breast cancers. Combining this analysis with analyses of KEGG cytokine pathway genes in the Richardson (GSE5460, N=129), Bild (GSE3143, N=158), Wang (GSE2034, N=287), Børrensen-Dale (GSE19783, N=115), and Chin Datasets (E-TABM-158, N=119), obtained from the Gene Expression Omnibus and ArrayExpress Repositories, we identified a consensus set of 58 inflammatory genes that are differentially expressed in ER-negative breast cancers as compared to ER+ or luminal breast cancers (genes were considered statistically significant if their p-value was less than 0.05 and displayed fold-differences of &amp;gt;1.5 or &amp;lt; 0.7 in at least two of the datasets analyzed). Further analysis of inflammatory gene expression in breast cancer cell lines revealed that 24 of these genes were highly expressed in basal-like cell lines at the RNA and protein level, as compared to luminal-like ER+ breast cancer lines (showing fold-differences of &amp;gt;1.5 or &amp;lt;0.7). To investigate the role these differentially-expressed genes play in regulating cell growth and transformation, we performed shRNA knock-down studies for 18 of these genes. These studies demonstrated that knockdown of many of these inflammatory genes (including IL-6, IL-8, CCL5, BMP2, TNFRSF21) reduced anchorage-independent growth of basal-like breast cancer cell lines. Further investigation of two cytokine genes, IL-6 and IL-8, revealed that autocrine tumor cell production of these cytokines was critical for maintenance of the transformed state of multiple basal-like cancer lines. We also found that combined inhibition of IL-6 and IL-8 using Tet-induced shRNA suppressed anchorage-independent growth more than either alone, suggesting that these cytokines have non-redundant roles in regulating autocrine-mediated cell growth. These findings suggest that these specific inflammatory pathways regulate the growth and maintenance of the transformed phenotype of basal breast cancers. These results suggest that combined inhibition of both IL-6 and IL-8 inflammatory pathways represents a promising strategy to treat or prevent basal breast cancers. These studies were supported by a Susan G. Komen Promise Grant (to PHB and GM). Citation Information: Cancer Prev Res 2011;4(10 Suppl):A42.

Analytic Study of Performance of Error Estimators for Linear Discriminant Analysis

We derive double asymptotic analytical expressions for the first moments, second moments, and cross-moments with the actual error for the resubstitution and leave-one-out error estimators in the case of linear discriminant analysis in the multivariate Gaussian model under the assumption of a common known covariance matrix and a fixed Mahalanobis distance as dimensionality approaches infinity. Sample sizes for the two classes need not be the same; they are only assumed to reach a fixed, but arbitrary, asymptotic ratio with the dimensionality. From the asymptotic moment representations, we directly obtain double asymptotic expressions for the bias, variance, and RMS of the error estimators. The asymptotic expressions presented here generally provide good small sample approximations, as demonstrated via numerical experiments. The applicability of the theoretical results is illustrated by finding the minimum sample size to bound the RMS in gene-expression classification.