Genome annotation

The genome annotation was limited to contigs containing at least 1 Kbp of non repeat
sequences and at least one putative complete gene based on alignment with spruce
transcriptomics sequences.
The genomes were annotated with MAKER2 pipeline (Holt and Yandel, 2011) which
incorporates AUGUSTUS (Stanke et al., 2006), SNAP (Korf, 2004) and GeneMark (Lomsadze
et al., 2005) gene predictors together with transcriptomics and proteomics evidence.
AUGUSTUS metaparameters were optimized with semi-automatic training protocol, the
evaluation was made by splitting into training and test sets. SNAP and AUGUSTUS were
trained with high quality gene models generated by a preliminary run of MAKER, and selected
by eAED score and QI tag. GeneMark was self trained to produce predictions.
Each species used a common input evidence from full-length cDNA (Rigault et al., 2011) (Ralph
et al., 2008) and SwissProt plant proteins. RNAseq transcriptomics assemblies were used for
Interior and Sitka spruce: short reads RNAseq libraries were assembled with pooled assembly
approach in RNA-Bloom v0.9.8 (Nip et al., XX). The transcripts were screened for contaminants
and only transcripts with complete CDS were used for annotation.
MAKER was run iteratively in two steps. The first run of MAKER included species-specific
evidence and the second run used a common data set evidence from the four genotypes.