1.1 Overview
   iODA, short for integrative Omics Data Analysis, is user-friendly Java based graphical interface software. It implements an integrative platform for selection of statistical methods for heterogeneous data and performing different algorithms to detect differentially expressed genes. Then users can perform the pathway enrichment analysis based on human KEGG pathways (http://www.genome.jp/kegg). The differentially expressed genes can be obtained by 3 kinds of omics data; i.e mRNA, microRNA and ChIP-Seq data. All the panels and functions integrated in iODA can be used individually to tackle the analysis process. Compared to other software based on integrative analysis, iODA is the first local, algorithm selection, cross-multiple level of datasets, pathway enrichment analysis and easy to use software in one integrative platform. Then we make a consistency analysis on the results by an overlapping analysis.
  We proposed six algorithms for heterogeneous detection incorporated in iODA are the list below.
  ●T-test - A t-test is any statistical hypothesis test in which the test statistic follows a Student's t distribution if the null hypothesis is supported. It can be used to determine if two sets of data are significantly different from each other, and is most commonly applied when the test statistic would follow a normal distribution if the value of a scaling term in the test statistic were known.
Recently, it has been recognized that many oncogenes show altered expression in only a small proportion of cancer samples(Lian, 2008). Such features will be removed when using t-test or t-test like methods because they average gene expression levels in all the studied samples. Tomlin et al. conclude that t-tests were not adequate for detecting heterogeneous patterns of oncogenes (Tomlins, et al., 2005).
  ●LSOSS - LSOSS is the abbreviation of Least Sum of Ordered Subset Squared, The general idea of LSOSS is to use the sum of squares of two ordered subsets of cancer samples to estimate the square sum of the t-statistic and to use the mean value of the appealing subset of cancer samples to estimate the mean value of cancer samples of the t-statistic(Wang and Rekaya, 2010).
  ●MOST - MOST is the abbreviation of maximum ordered subset t-statistics, Lian et al proposed another statistics for the detection of cancer differential gene expression which have similar power to ORT when the number of activated samples is very small, but perform better when more samples are differentially expressed(Lian, 2008).
  ●COPA - Tomlins et al (2005) have proposed the "cancer outlier profile analysis" (COPA) method for detecting cancer genes which show increased expressions in a subset of disease samples. They argue that in the majority of cancer types, oncogene has heterogeneous activation patterns; traditional analytical methods, for example, t-statistic, which search for common activation of genes across a class of cancer samples, will fail to find such oncogene expression profiles(Wu, 2007).
   ●ORT - Wu et al study statistical methods to detect cancer genes that are over- or down-expressed in some but not all samples in a disease group. This has proven useful in cancer studies where oncogenes are activated only in a small subset of samples. They propose the outlier robust t-statistic (ORT), which is intuitively motivated from the t-statistic; the most commonly used differential gene expression detection method(Wu, 2007).
  ●OS - Tibshirani et al proposed a method for detecting genes that, in a disease group, exhibit unusually high gene expression in some but not all samples. This can be particularly useful in cancer studies, where mutations that can amplify or turn off gene expression often occur in only a minority of samples.
  We propose three different omics level data which can be analyzed in iODA are the list below:
  1.Gene expression data (mRNA level)
  2.MicroRNA level data
  3.ChIP-Seq level data
  iODA provide a novel platform with emphasis on the pathways consistency analysis generated from pathway enrichment analysis from multiple omics data, which will finally lead to a robust pathway list. This software is particularly useful for researchers who want to integrate different level omics data and various datasets from different researchers for further pathway enrichment analysis and make a consistency analysis on the pathway results. We have verified the hypothesis that common molecular signatures are more similar at the pathway level than at the gene level. So we propose this platform to eliminate the heterogeneity of different datasets and integrative the different omics datasets at a system level and obtain deeper insight on the underlying biological mechanisms of the data generated from gene microarrays, microRNA microarrays and next generation sequencing ChIP-Seq data.
  With the application of software iODA, biologists will find it easy to analyze the omics data from different omics levels and datasets with pathway enrichment analysis. The whole workflow for each dataset analysis was integrated in iODA, such as test for best algorithm to perform outlier detection, differentially expressed genes detection by appropriate algorithm, differentially expressed microRNAs detection by appropriate algorithm, microRNAs target genes detection, ChIP-Seq data peaks detection, KEGG pathway enrichment analysis, pathway consistency analysis with different omics level.
  This tutorial helps users to run iODA on Linux (Ubuntu for example) and Windows (7 or later versions) to facilitate the biological data interpretation in a step-by-step style.

1.2 Basic concepts
1.2.1 Pathway
  A pathway is a group of related, genes, metabolites, and their mutual interactions, which form an aggregate biological function. A similar concept is gene/compound set, which means some manually defined gene/compound sets according to their functionality or ontology. Gene/compound set can be interpreted as a superior concept of pathway.
  The KEGG pathway map is a molecular interaction/reaction network diagram represented in terms of the KEGG Orthology (KO) groups, so that experimental evidence in specific organisms can be generalized to other organisms through genomic information. Each map is manually drawn with in-house software called KegSketch, which generates the KGML+ file. This file is an SVG file containing graphics objects that are associated with KEGG objects.
1.2.2 Gene expression data
  Gene expression is the process by which information from a gene is used in the synthesis of a functional gene product. In genetics, gene expression is the most fundamental level at which the genotype gives rise to the phenotype. The genetic code stored in DNA is "interpreted" by gene expression, and the properties of the expression give rise to the organism's phenotype. Such phenotypes are often expressed by the synthesis of proteins that control the organism's shape, or that act as enzymes catalyzing specific metabolic pathways characterizing the organism.
  Publicly available microarray expression datasets can be downloaded by Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) database which had been generated by independent laboratory. These datasets were measured with different technologies and platforms. Comparative cancer analyses included cancer versus respective normal tissue, high grade versus low grade cancer, poor outcome versus good outcome, metastatic versus primary cancer, and subtype1 versus subtype2. Thus, our analysis across multiple datasets, based on normal prostate versus tumor prostate samples, was comparable.
  1.2.3 MicroRNA data
  MicroRNAs (miRNAs) are small non-coding RNAs of approximately 22-nucleotides. They play important roles in gene regulation at post-transcriptional level. They are able to repress the activity of complementary mRNAs by targeting the 3’-untranslated regions(Bartel, 2009). Release 19 of the miRBase database contains more than 2200 mature miRNA sequences for human(Kozomara and Griffiths-Jones, 2011). Aberrant miRNA expression was shown related to the generation of cancer stem cells and the tumor genesis(Lin, et al., 2010; Liu, et al., 2011; Mallick, et al., 2011). Microarray-based technologies have routinely been used for profiling molecular expression in cancer. Microarray allows simultaneous expression profiling of tens of thousands of genes in normal versus malignant cells. We can use these miRNA microarray datasets to detect the differentially expressed miRNAs. Then we obtain the target genes of them.
1.2.4 ChIP-Seq data
  The disease associated ChIP-Seq datasets were extracted from Gene Expression Omnibus (GEO). Peak detection algorithm is crucial to the analysis of ChIP-Seq dataset(Ding, et al., 2012). Currently, several tools are available to identify genome-wide binding sites of transcription factors, such as FindPeaks(Fejes, et al., 2008), F-Seq (Boyle, et al., 2008), CisGenome(Ji, et al., 2011), MACS (Zhang, et al., 2008), SISSRs (Narlikar and Jothi, 2012), and QuEST (Valouev, et al., 2008). These different methods have their own advantages and disadvantages, although they act in a similar manner. ChIP-Seq data has regional biases because of sequencing and mapping biases, chromatin structure, and genome copy number variations (Redon, et al., 2006). In order to get more stable result, the iODA integrate MACS tool to identify the binding sites of disease related transcription factors in this study. The tool uses control samples to guide peak finding and calculate the FDR (False Discovery Rate) value of peaks. Then iODA also provide the PeakAnalyzer tool to annotate the peaks.
1.2.5 Pathway enrichment analysis
  Pathway Enrichment Analysis is the analysis to find the most relevant pathway, according to the gene/protein data, by calculating their differentially expression (DE) values to fit in an enrichment model, and finally returns a list of pathways ranked by certain enrichment score. Currently, most evaluations of enrichment score employ probability approach, such as p-value.
  Pathway enrichment analysis helps reveal the underlying biological pathway/function from large biological data. It is the most commonly used in-silico function analysis for high-throughput omics data.
  In iODA, KEGG, an open access database resource, was selected for pathway enrichment analysis. Both “Local pathway enrichment” and “Online pathway enrichment” functions were designed. The former utilizes the local KEGG database and calculates p-values through hypergeometric distribution algorithm, whereas the latter uses HTTP protocol to submit a request to the KEGG web server and receives the pathway results. Here the local pathway database needs to be periodically updated and the online function requires an internet connection, so the response time is based on the status of network and the KEGG server. After submitting the dysfunctional gene list, the returning result contains information of enriched pathway ID, pathway symbol, the number of submitted genes in this pathway, the number of all genes in this pathway, p-value, q-value and DE-gene list.
1.3 Pipeline
  iODA guides users to follow the pipeline, i.e., selection of statistical methods for heterogeneous data, expression data input, detect the differentially expressed outliers with the selected algorithm, microRNA target genes detection, ChIP-Seq data annotation, pathway enrichment analysis, cross mapping and consistency analysis to perform meta-analysis on pathway results, (Figure 1).
  In this study, iODA integrative three different levels of omics data and make a meta-analysis on the pathway results. For each omics level, there are a few datasets from the different laboratories, thus we make a consistency analysis on pathway level to get the common pathway results. Then we make a meta-analysis for all three omics level results.

Fig. 1. The working pipeline of iODA. It provides the function of DE-gene/miRNA identification, pathway enrichment and consistency analysis.

Fig. 5. The interface after clicking the "Add dataset" button

Fig. 6.  The dataset list after three datasets are added.
4.3 Selection of statistical methods
  After importing the test datasets, users click the button "Run test" to test for the best algorithm. A bar chart of the result will give an indicative of the best algorithm selection. An example of bar chart shows below. The different color means different dataset, the Y-axis means the percentage of the putative outliers in each dataset.

Fig. 7. Overlapping percentages of putative outliers. The different colors signify different datasets, the X-axis represents the six algorithms, and the Y-axis indicates the percentage of the putative outliers in each dataset.
The putative outlier percentages are calculated as follows:
1. Detecting the outliers of each dataset using the six algorithms (termed as Oi, 1=< i <=6);
2. Counting the number of occurrences of each outlier in the six outlier results from step 1;
3. Screening the outliers with number of occurrences no fewer than four (termed as OF);
4. The accuracy percentage is calculated as OF / Oi (1=< i <=6)