Resistance gene detection was assessed using a cutoff of ≥90% coverage and ≥90% identity to define the presence of a gene.
There were three steps for identifying shared single copy genes: (1) all protein data sets were compared against each other using an all-blast-all BLAST with an e-value cutoff of 1e-5, coferate of at least 70% of the query protein and identity of at least 30%.
The BLASTCLUST program [ 63] set up with a length coverage cutoff of 0.8 and a score coverage threshold (bit score divided by alignment length) of 0.5 was used for clustering of B. massiliensis proteins to identify expanded families.
To incorporate more evolutionarily distant relationships and connect genes in singleton groups to more distantly related homologs, unconnected genes were assigned to the module containing the gene with the most similar protein sequence if its best match exceeded a less stringent cutoff of 10−10 and coverage of at least 50%.
We used a cutoff of 30%% genome coverage for reliable detection of expressed viruses [Bukur et al., manuscript in preparation].
With a stringent cutoff of ≥ 10X coverage in all clones, we acquired data on 25.62Mb of the genome, whereas 64.09Mb of the genome was covered at ≥ 10X in at least one clone.
The sequences of genes and unigenes thus obtained, as well as those of the predicted carnation genes, were mapped onto the KEGG reference pathways by BLAST searches against genes in the KEGG database with E-value cutoff of 1 E–10, length coverage ≥25% and identity ≥50%, and the status of mapping was compared among the four plant species.
We used Velvet 1.0.12 (Zerbino and Birney 2008) and Velvet Optimiser (http://bioinformatics.net.au/software.velvetoptimiser.shtml, last accessed July 4, 2013) with a hash length of 71, expected coverage of 31, and coverage cutoff of 0.326 to perform de novo assembly of the mitochondrial genome of A. syriaca.
The t7 phage metagenomic data was quality trimmed using a Phred [39], [40] Q20 score and the reads were assembled using Velvet 1.0.05 [41] with a K-mer length of 37, expected coverage of 25 and a coverage cutoff of 2, which produced assemblies containing the longest contigs.
The Velvet parameters were hash length of 31, expected coverage of 104, and a coverage cutoff of 20.
