5.3　Annotation methods

Protein coding gene models were predicted using GeneMark-ES［42］with the default parameters and our own designed algorithm，based on homology to known genes.Glean［43］was used to get the best models from the two datasets.

The functionally annotation of predicted gene models were mainly based on homology to known annotated genes and BLAST was the mainly used tool in our analyses.We aligned all protein models by BLASTP to SwissProt，TrEMBL，and nr，InterPro，Pfam［44］and also mapped them onto functional terms，including GO［45］，COGs［46］and KEGG pathways［47］(BLASTP cut-off e-value＜1e-7).Since each gene mapped to different database sequences，there might be multiple aligned results meeting the cut-off，the annotations of the sequences with the best score were chosen to be the annotation of the gene in G.lucidum.(https://www.daowen.com)

Transposons were identified by aligning the assembled results with the known sequences of the transposon library.The specific method was made through the Repeat-Masker software(http：//www.repeatmasker.org，using Repbase database［48］)，Repeat Protein Masker software(using the transposon protein library that comes with Repeat Masker)and 3 other tools LTR-FINDER［49］，RepeatScout［50］and PILER［51］with default parameters.TRF(Tandem Repeat Finder)［52］software was used to predict tandem repeats.

rRNAs were identified by BLAST against the rRNA libraries or predicted by using rRNAmmer［53］software.tRNAscan-SE［54］software was used to detect tRNA regions and its secondary structures.Other non-coding RNAs such as miRNA，sRNA and snRNA were predicted by Rfam.