Fig. 1

Pipeline of siRNA analysis from prediction to statistical analysis. Data obtained from the SOLiD sequencing of the 12 libraries were cleaned of adaptors and special sequences (snRNA, snoRNA, mitRNA, tRNA, miRNA and pre-miRNA). The Shortstack algorithm mapped and identified clusters corresponding to genomic regions accumulating siRNAs. The algorithm was first run for each library independently. Bedtools was used to identify clusters that were present in at least two out of three libraries in at least one condition (galls or roots). If these clusters were separated by a distance of less than 2 nucleotides, they were then merged (default parameter) and selected to build a reference set of clusters to perform counting and statistical analysis. From counting data, a DESeq statistical analysis was performed to identify clusters that were differentially expressed between gall and root conditions. Only clusters with a DicerCall (DC-clusters) and a minimum coverage of 2 rpmm in all replicates in at least one condition (galls or roots) were considered as biologically relevant