RankProdIt: A web-interactive Rank Products analysis tool
© Laing et al; licensee BioMed Central Ltd. 2010
Received: 27 May 2010
Accepted: 6 August 2010
Published: 6 August 2010
The first objective of a DNA microarray experiment is typically to generate a list of genes or probes that are found to be differentially expressed or represented (in the case of comparative genomic hybridizations and/or copy number variation) between two conditions or strains. Rank Products analysis comprises a robust algorithm for deriving such lists from microarray experiments that comprise small numbers of replicates, for example, less than the number required for the commonly used t-test. Currently, users wishing to apply Rank Products analysis to their own microarray data sets have been restricted to the use of command line-based software which can limit its usage within the biological community.
Here we have developed a web interface to existing Rank Products analysis tools allowing users to quickly process their data in an intuitive and step-wise manner to obtain the respective Rank Product or Rank Sum, probability of false prediction and p-values in a downloadable file.
The online interactive Rank Products analysis tool RankProdIt, for analysis of any data set containing measurements for multiple replicated conditions, is available at: http://strep-microarray.sbs.surrey.ac.uk/RankProducts
The identification of differentially expressed or represented entities (genes/probes) between two conditions or strains, respectively, in a DNA microarray experiment is often the first task following data normalisation. However, to identify such entities it is no longer considered acceptable to apply an arbitrary fold-change threshold above which the difference in transcriptional or presence/absence status of an entity is defined. Instead, confidence through a test-statistic is expected. Of the many statistical methods that exist (and must be chosen between) for calculating test-statistics in the microarray field most are variants of the t-test, either traditional of modified. Whilst these methods are powerful their use has been shown to be limited when applied to 'noisy' data sets: few (less than 10) biological replicates and/or a high degree of variability between biological replicates [1, 2]. Rank Products analysis (and the similar approach of Rank Sum  analysis) is an algorithm with which a confidence value can be obtained and an alternative to those statistical methods that require many biological replicates with little variability; it is robust against the noise in a microarray experiment and still retains sensitivity [1–4].
To date users wishing to conduct Rank Products analysis on their own data set have had the options of 1) manually calculating the Rank Products and associated statistics, 2) using the R [5, 6] package RankProd  or 3) running a Perl script on their own computer . Clearly, option 1 is unsuitable due to the time it takes to prepare the data set, learn the protocol/algorithm and perform many calculations. Although option 2 avoids manual calculation it requires familiarity with R [5, 6] which can be daunting to some biologists. Whilst option 3 is less demanding (only requiring a command line entry) the script is only able to take log ratios and not linear data, which may therefore require a re-scaling of data. Thus, it is apparent that there is a need for an interface to the Rank Products tool such that users can analyse their own data in an intuitive, 'clickable' manner. Here we present the first online interactive Rank Products analysis tool RankProdIt.
RankProdIt is a web interface developed in haXe  that calls Perl CGI scripts to upload the data file, generate the R [5, 6] commands and execute R on the server in slave mode. For the Rank Products and Rank Sum analysis all user selected parameters are passed to the R package RankProd . Note, that the default 100 permutations of RankProd, for calculating the probability of observing a Rank Product and/or Rank Sum by chance, for both Rank Products and Rank Sum analysis is retained in RankProdIt.
RankProdIt is a generic tool, able to accept any data set containing replicated samples for at least two conditions. Thus, whilst this manuscript documents RankProdIt for microarray data analysis, it can be applied to other high throughput data sets such as next-generation sequencing, proteomic and metabolomic data. Input measurements can either be in the form of absolute levels, where row-element k has measurements in multiple columns for each condition i and j, such as that obtained from single-colour microarray experiments, or in the form of ratios, where each column of row-element k is a ratio of conditions (i/j), as obtained from two-colour microarray experiments.
To process data using RankProdIt a user submits a tab-delimited text file that contains a row-identifier (typically gene/probe identifier) column and several columns containing data; missing data is represented by NA or NaN. The input file is not required to have columns in any particular order and columns containing data not to be used in the analysis can also be included. A header row does not need to be included but if so, there must only be one. An example input file is given in Additional file 1.
at least two columns as condition 1 samples and at least two columns as condition 2 samples
at least two columns as condition1/condition2 or condition2/condition1 samples
If the correct selections are not made an error message is given following submission. Note, that whilst it is possible to perform Rank Products and/or Rank Sum analysis with as few as two biological replicates for each condition, it is recommended that a greater number of replicates be provided for greater confidence in data reliability.
The scale of the input data and the presence of a column header row is automatically selected by RankProdIt. Prior to submission the user can select whether to perform Rank Products or Rank Sum analysis; by default Rank Products analysis is selected.
Upon successful submission the data selected by the user is imported into R and Rank Products or Rank Sum analysis is performed by the RankProd package ; whilst the Rank Products/Sum analysis is being conducted an indication of processing is given, alerting the user that the analysis has not finished. If the data and selections made by the user do not cause an error within the RankProd  package a link to the output file is provided, for the user to download the results.
An example of an output (results) file is given in Additional file 2 and a brief description of columns within an output file is provided in Additional file 3. The output tab-delimited text file of RankProdIt is suitable to open with any spreadsheet software for data interpretation and/or further analysis (e.g. the enhanced distribution calculations for Rank Products that can easily be calculated in Excel ).
Availability and requirements
Project name: RankProdIt
Project home page: http://strep-microarray.sbs.surrey.ac.uk/RankProducts/
Any restrictions to use by non-academics: None
This work was funded by the European Commission's FP6 programme: ActinoGEN, contract no. 005224 (awarded to CPS).
- Jeffery IB, Desmond GH, Culhane AC: Comparison and evaluation of methods for generating differentially expressed gene lists from microarray data. BMC Bioinformatics. 2006, 7: 359-10.1186/1471-2105-7-359.PubMed CentralPubMedView ArticleGoogle Scholar
- Hong F, Breitling R: A comparison of meta-analysis methods for detecting differentially expressed genes in microarray experiments. Bioinformatics. 2008, 24: 374-382. 10.1093/bioinformatics/btm620.PubMedView ArticleGoogle Scholar
- Breitling R, Herzyk P: Rank-based methods as a non-parametric alternative of the T-statistic for the analysis of biological microarray data. J Bioinform Comput Biol. 2005, 3: 1171-1189. 10.1142/S0219720005001442.PubMedView ArticleGoogle Scholar
- Breitling R, Armengaud P, Amtmann A, Herzyk P: Rank products: a simple, yet powerful, new method to detect differentially regulated genes in replicated microarray experiments. FEBS Letters. 2004, 573: 83-92. 10.1016/j.febslet.2004.07.055.PubMedView ArticleGoogle Scholar
- R. [http://www.R-project.org]
- R Development Core Team: R: A language and environment for statistical computing. 2005, R Foundation for Statistical Computing:. Vienna, Austria, ISBN 3-900051-07-0.Google Scholar
- Hong F, Breitling R, McEntee CW, Wittner BS, Nemhauser JL, Chory J: RankProd: A Bioconductor package for detecting differentially expressed genes in meta-analysis. Bioinformatics. 2006, 22: 2825-2827. 10.1093/bioinformatics/btl476.PubMedView ArticleGoogle Scholar
- GlaMA. [http://www.brc.dcs.gla.ac.uk/systems/glama/]
- haXe. [http://haxe.org/]
- Koziol JA: Comment son the rank product method for analyzing replicated experiments. FEBS Lett. 2010, 584: 941-944. 10.1016/j.febslet.2010.01.031.PubMed CentralPubMedView ArticleGoogle Scholar