• The Korean Journal of Applied Statistics
• /
• v.29 no.3
• /
• pp.437-448
• /
• 2016

Gene set analysis utilizing biologic information is expected to produce more interpretable results because the occurrence of tumors (or diseases) is believed to be associated with the regulation of related genes. Many methods have been developed to identify statistically significant gene sets across different phenotypes; however, most focus exclusively on either the differential gene expression or the differential correlation structure in the gene set. This research provides a new method that simultaneously considers the differential expression of genes and differential coexpression with multiple genes in the gene set. Application of this NEW method is illustrated with real microarray data example, p53; subsequently, a simulation study compares its type I error rate and power with GSEA, SAMGS, GSCA and GSNCA.

• Communications for Statistical Applications and Methods
• /
• v.16 no.3
• /
• pp.397-408
• /
• 2009

The development of microarray technology makes possible to analyse many thousands of genes simultaneously. While it is important to test each gene whether it shows changes in expression associated with a phenotype, human diseases are thought to occur through the interactions of multiple genes within a same functional cafe-gory. Recent research interests aims to directly test the behavior of sets of functionally related genes, instead of focusing on single genes. Gene set enrichment analysis(GSEA), significance analysis of microarray to gene-set analysis(SAM-GS) and parametric analysis of gene set enrichment(PAGE) have been applied widely as a tool for gene-set analyses. We describe their problems and propose an alternative method using a parametric analysis by adopting normal score transformation of gene expression values. Performance of the newly derived method is compared with previous methods on three real microarray datasets.

• BMB Reports
• /
• v.47 no.6
• /
• pp.348-353
• /
• 2014

The speed of sound (SOS) value is an indicator of bone mineral density (BMD). Previous genome-wide association (GWA) studies have identified a number of genes, whose variations may affect BMD levels. However, their biological implications have been elusive. We re-analyzed the GWA study dataset for the SOS values in skeletal sites of 4,659 Korean women, using a gene-set analysis software, GSA-SNP. We identified 10 common representative GO terms, and 17 candidate genes between these two traits (PGS < 0.05). Implication of these GO terms and genes in the bone mechanism is well supported by the literature survey. Interestingly, the significance levels of some member genes were inversely related, in several gene-sets that were shared between two skeletal sites. This implies that biological process, rather than SNP or gene, is the substantial unit of genetic association for SOS in bone. In conclusion, our findings may provide new insights into the biological mechanisms for BMD.

• Molecules and Cells
• /
• v.24 no.2
• /
• pp.200-209
• /
• 2007

We generated gene expression data from the tissues of 50 gastric cancer patients, and applied meta-analysis and gene set analysis to this data and three other stomach cancer gene expression data sets to define the gene expression changes in gastric tumors. By meta-analysis we identified genes consistently changed in gastric carcinomas, while gene set analysis revealed consistently changed biological themes. Genes and gene sets involved in digestion, fatty acid metabolism, and ion transport were consistently down-regulated in gastric carcinomas, while those involved in cellular proliferation, cell cycle, and DNA replication were consistently up-regulated. We also found significant differences between the genes and gene sets expressed in diffuse and intestinal type gastric carcinoma. By gene set analysis of cytogenetic bands, we identified many chromosomal regions with possible gross chromosomal changes (amplifications or deletions). Similar analysis of transcription factor binding sites (TFBSs), revealed transcription factors that may have caused the observed gene expression changes in gastric carcinomas, and we confirmed the overexpression of one of these, E2F1, in many gastric carcinomas by tissue array and immunohistochemistry. We have incorporated the results of our meta- and gene set analyses into a web accessible database (http://human-genome.kribb.re.kr/stomach/).

• The Korean Journal of Applied Statistics
• /
• v.31 no.2
• /
• pp.241-251
• /
• 2018

The analysis of gene sets in microarray has advantages in interpreting biological functions and increasing statistical powers. Many statistical methods have been proposed for detecting significant gene sets that show relations between genes and phenotypes, but there is no consensus about which is the best to perform gene sets analysis and permutation based tests are considered as standard tools. When many gene sets are tested simultaneously, a large number of random permutations are needed for multiple testing with a high computational cost. In this paper, several parametric approximations are considered as alternatives of the permutation distribution and the moment based gene set test has shown the best performance for providing p-values of the permutation test closely and quickly on a general framework.

• Journal of the Korean Data and Information Science Society
• /
• v.23 no.1
• /
• pp.1-11
• /
• 2012

When the microarray experiment developed, main interest was limited to detect differentially expressed genes associated with a phenotype of interest. However, as human diseases are thought to occur through the interactions of multiple genes within a same functional category, the unit of analysis of the microarray experiment expanded to the set of genes. For the phenotype of censored survival time, Gene Set Enrichment Analysis(GSEA), Global test and Wald type test are widely used. In this paper, we modified the Wald type test by adopting normal score transformation of gene expression values and developed a parametric test which requires much less computation than others. The proposed method is compared with other methods using a real data set of ovarian cancer and a simulation data set.

• Asian Pacific Journal of Cancer Prevention
• /
• v.14 no.3
• /
• pp.1629-1633
• /
• 2013

Background: Gene set analysis (GSA) incorporates biological with statistical knowledge to identify gene sets which are differentially expressed that between two or more phenotypes. Materials and Methods: In this paper gene sets differentially expressed between acute lymphoblastic leukaemia (ALL) with BCR-ABL and those with no observed cytogenetic abnormalities were determined by GSA methods. The BCR-ABL is an abnormal gene found in some people with ALL. Results: The results of two GSAs showed that the Category test identified 30 gene sets differentially expressed between two phenotypes, while the Hotelling's $T^2$ could discover just 19 gene sets. On the other hand, assessment of common genes among significant gene sets showed that there were high agreement between the results of GSA and the findings of biologists. In addition, the performance of these methods was compared by simulated and ALL data. Conclusions: The results on simulated data indicated decrease in the type I error rate and increase the power in multivariate (Hotelling's $T^2$) test as increasing the correlation between gene pairs in contrast to the univariate (Category) test.

• Asian Pacific Journal of Cancer Prevention
• /
• v.17 no.2
• /
• pp.835-838
• /
• 2016

Background: Lung cancer remains one of the most common cancers in the world, both in terms of new cases (about 13% of total per year) and deaths (nearly one cancer death in five), because of the high case fatality. Errors in lung cancer type or malignant growth determination lead to degraded treatment efficacy, because anticancer strategy depends on tumor morphology. Materials and Methods: We have made an attempt to evaluate effectiveness of machine learning algorithms in the task of lung cancer classification based on gene expression levels. We processed four publicly available data sets. The Dana-Farber Cancer Institute data set contains 203 samples and the task was to classify four cancer types and sound tissue samples. With the University of Michigan data set of 96 samples, the task was to execute a binary classification of adenocarcinoma and non-neoplastic tissues. The University of Toronto data set contains 39 samples and the task was to detect recurrence, while with the Brigham and Women's Hospital data set of 181 samples it was to make a binary classification of malignant pleural mesothelioma and adenocarcinoma. We used the k-nearest neighbor algorithm (k=1, k=5, k=10), naive Bayes classifier with assumption of both a normal distribution of attributes and a distribution through histograms, support vector machine and C4.5 decision tree. Effectiveness of machine learning algorithms was evaluated with the Matthews correlation coefficient. Results: The support vector machine method showed best results among data sets from the Dana-Farber Cancer Institute and Brigham and Women's Hospital. All algorithms with the exception of the C4.5 decision tree showed maximum potential effectiveness in the University of Michigan data set. However, the C4.5 decision tree showed best results for the University of Toronto data set. Conclusions: Machine learning algorithms can be used for lung cancer morphology classification and similar tasks based on gene expression level evaluation.

• Communications for Statistical Applications and Methods
• /
• v.29 no.5
• /
• pp.591-601
• /
• 2022

In gene set analysis with microarray expression data, a group of genes such as a gene regulatory pathway and a signaling pathway is often tested if there exists either differentially expressed (DE) or differentially co-expressed (DC) genes between two biological conditions. Recently, a statistical test based on covariance estimation have been proposed in order to identify DC genes. In particular, covariance regularization by hard thresholding indeed improved the power of the test when the proportion of DC genes within a biological pathway is relatively small. In this article, we compare covariance thresholding methods using four different regularization penalties such as lasso, hard, smoothly clipped absolute deviation (SCAD), and minimax concave plus (MCP) penalties. In our extensive simulation studies, we found that both SCAD and MCP thresholding methods can outperform the hard thresholding method when the proportion of DC genes is extremely small and the number of genes in a biological pathway is much greater than a sample size. We also applied four thresholding methods to 3 different microarray gene expression data sets related with mutant p53 transcriptional activity, and epithelium and stroma breast cancer to compare genetic pathways identified by each method.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.47 no.5
• /
• pp.17-24
• /
• 2010

Recently the study of identifying bio-markers for disease diagnosis and prognosis has been actively performed. In particular, lots of attentions have been paid to the finding of pathway gene-sets differentially expressed in disease patients rather than the finding of individual gene markers. In this paper we propose a novel method to identify disease-related pathway gene-sets based on time-series microarray data. For this purpose, we firstly compute individual gene scores by the using maSigPro (microarray Significant Profiles) and then arrange all the genes in the decreasing order of the corresponding gene scores. The rank of each gene in the entire list is used to evaluate the statistical significance of candidate gene-sets with Wilcoxson rank sum test. For the generation of candidate gene-sets, MSigDB (Molecular Signatures Database) pathway information has been employed. The experiment was conducted with prostate cancer time-series microarray data and the results showed the usefulness of the proposed method by correctly identifying 6 out of 7 biological pathways already known as being actually related to prostate cancer.