Merging R into the web: DNMAD & Tnasas, web-based resources for microarray data analysis.

DNMAD and Tnasas are two web-based resources for microarray data analysis (normalization and class prediction), that implement R code into a web server conflguration. These tools are included in GEPAS, a suite for gene expression data analysis [4]. The flrst tool we present is called DNMAD that stands for \Diagnosis and Normalization for MicroArray Data". DNMAD is essentially a web-based interface for the Bioconductor package limma [7]. DNMAD has been designed to provide a simple and at the same time robust and fast way for normalization of microarray data. We provide global and print-tip loess normalization [9]. We also ofier options for the use of quality ∞ags, background adjustment, input of compressed flles with many of arrays, and slide-scale normalization. Finally, to enable a fully featured pipeline of analysis, direct links to the central hub of GEPAS, the Preprocessor [5] are provided. The tool, help flles, and tutorials are available at http://dnmad.bioinfo.cnio.es. The second tool we present is called Tnasas for \This is not a substitute for a statistician". Tnasas performs class prediction (using either SVM, KNN, DLDA, or Random Forest) together with gene selection, including cross-validation of the gene selection process to account for \selection bias" [1], and cross-validation of the process of selecting the number of genes that yields the smallest error. This tool has mainly an exploratory and pedagogical purpose: to make users aware of the widespread problem of selection bias and to provide a benchmark against some (overly) optimistic claims that occasionally are attached to new methods and algorithms; Tnasas can also be used as a \quick and dirty" way of building a predictor. Tnasas uses packages e1071 [2], randomForest [6], and class (part of the VR bundle [8]). The tool, help flles, and tutorials are available at http://tnasas.bioinfo.cnio.es. Both tools use a Perl-CGI script that checks all the data input by the user and, once checked, sends it to an R script that performs the actions requested by the user. To merge the Perl-CGI script with the script in R we use the R package CGIwithR [3]. Source code for both applications will be released under the GNU GPL license.