ACT representative primary microarray samples

Arabidopsis thaliana ATH1 raw microarray data (CEL files) were download from ArrayExpress, GEO and NASCArrays. After duplicate and corrupt sample removal, using an in-house PHP script, our dataset consisted of 19,887 unique microarray samples from 1390 studies. Finally, after quality control, 6933 distinct, wild-type, healthy samples were selected. Pairwise sample correlations were calculated using Pearson Correlation Coefficient, using the expression values of 21273 non-obsolete Arabidopsis thaliana genes, in the 6933 previously selected samples. A sample distance matrix was created using the d = 1 – r formula and a sample correlation tree of 6933 sample-leaves was created in Newick format, based on the distance matrix. Then, the tree of 6933 sample-leaves, was programmatically pruned in an iterative procedure using an in-house algorithm trimming close leaves, where in each iteration the leaf with the shortest distance to its first common node was trimmed, leaving 3500 leaves which represent the most distinct samples.