To maximize the statistical reliability of the data, three biological replicates were carried out. In addition, for each time
point comparison and each biological replicate, three technical replicates (cDNA obtained from the same mRNA extraction) were used for hybridization. For one of the three technical replicates, the labelling of the two cDNA samples with either Cy5 or Cy3 fluorescent dye was reversed to prevent potential dye-related differences in labelling efficiency. Overall, 27 images were analysed, 9 for each time point during Xoo infection. The nine data points obtained for each gene were used in the analyses. Microarray data analysis The slides were scanned, using a chip reader/scanner (Virtek Vision International, Inc., Waterloo, ON, BV-6 supplier Canada). The signal was initially normalized during image buy GANT61 scanning to adjust the average ratio between the two channels, using control spots. Spot intensities from scanned slides were quantified, using the Array-Pro 4.0 software
(Media Cybernetics, Inc., Silver Spring, MD, USA). With this program, local corner background correction was carried out. Array-Pro 4.0 output data files (in Excel) were used to perform the lowest intensity normalization, standard deviation regularization, low intensity filtering, and dye-swap analysis, using the MIDAS computer program [68]. Normalization between different slides was carried out by centring [69]. MIDAS [68] was also used for replicate analysis and dye-swap filtering. Bootstrap analyses with SAM enabled us to identify the differentially expressed genes, using Diflunisal a cut-off of two and adjusting the delta-delta Ct value, FDR, and FSN to minimize the number LDN-193189 ic50 of false positives genes [70]. We conducted k-means clustering analysis to group the cDNA clones according to the similarity of their expression patterns, using MeV software available from TIGR and the default
options [68]. Sequence data analysis The 710 genes identified as differentially expressed were one-end sequenced. Sequence data were processed, using a PerlScript pipeline, to remove vector and low-quality sequences and to assemble sequences into a non-redundant set of sequences [71]. The Xoo MAI1 non-redundant set of sequences was deposited at GenBank’s GSS Database http://www.ncbi.nlm.nih.gov/dbGSS/[72], under accession numbers FI978231-FI978329. Processed sequences were initially searched against the NCBI database with BLASTN and TBLASTX http://blast.ncbi.nlm.nih.gov/Blast.cgi[73], setting BLAST parameters to search against the complete non-redundant database and the genomes of Xoo strains KACC10331, MAFF311018, and PXO99A, and Xoc strain BLS256. A BLAST search was also performed with the partial genome of the African Xoo strain BAI3, which is currently being sequenced (Genoscope project 154/AP 2006-2007 and our laboratory, 2009, unpublished data). Results of these comparisons are summarized in the Additional file 1, Table S1.