DATAN, a software package based on the proprietary technology of the Swedish company MultiD, was developed to decompose spectroscopic data into fundamental components. The adaptation of DATAN to real-time PCR presents a fresh perspective into options for data analysis. DATAN’s aim is to help the users extract biological significance from their data. For instance, DATAN’s hierarchical agglomerate clustering function lets the user recognize genes with similar expression profiles or cluster samples according to their profile (e.g., classification of samples to sets of healthy and sick patients). The other functions of DATAN, Principle Component Analysis (PCA) and Neural Network (NN), essentially serve the same purpose. What may look like redundancy is actually an advantage of DATAN, in terms of providing a tool to test clustering questions using different approaches and relying on different heuristics. The multiple paths permitted by DATAN enable more robust results. In essence, an expression pattern recognized using alternative methods are more likely to represent real biological phenomena, not a result of analysis model local heuristics.
It is easy to upload files to DATAN in various formats (Excel spreadsheets, text and mdf) and the explanation for editing data before uploading is simple and clear. DATAN does allow data editing, but the icons are new and may not be easily recognized, initially (easily solved by editing data in Excel prior to analysis). The weakness of DATAN is in data presentation. No options exist for naming axes or displaying units/titles in PCA, dendrogram or GeNorm plots. This wouldn’t be a problem if DATAN would have had an ‘Export’ function, enabling graphical editing of the results on other software. Unfortunately, DATAN lacks this important option.
In spite of the weaknesses, the multiple options for clustering in DATAN make it a good tool for pattern recognition. To make the most of DATAN, the user should have a basic understanding of the analysis methods (assistance provided in the “Help” function of the software). For example, the algorithm GeNorm finds the optimal reference genes for normalization of gene expression by selecting the genes with the most similar expression pattern, thus, it may be inaccurate once co-regulated genes are in the tested set of genes. Being familiar with the model’s heuristic will probably motivate the user to verify the results with additional methods based on other heuristics (e.g NormFinder).
The inclusion of a step-by-step wizard to explain how/why to set each parameter would have eased initial software confusion. The need for guidance is particularly relevant in setting the parameters of the Neural Network.
DATAN is a useful and inexpensive tool for pattern recognition and may be a reasonable solution in labs that can not afford an expensive statistics package. The pitfall of DATAN lies in its poor options to set presenation of results and lack of an export function.
Tzachi Bar
M.Sc.
Department of Molecular Biotechnology
Chalmers University