Both facts based and ab initio gene models were used for annotation. Transcriptomes of fighting fish and protein sequences of zebrafish, medaka, stickleback, fugu, and Nile tilapia from Ensembl database unlock 86 were used for proof. SNAP Korf 2004 and Augustus Stanke and Waack 2003 were iteratively used for ab initio gene models training. Predicted protein sequences were annotated by blast to nr and RefSeq databases Pruitt et al. Identification of CNEs was in response to a prior method Brawand et al. 2014.