Tissues for transcriptome analyses were obtained from a homozygous clonal 1-year-old female sampled 3 weeks after spawning. These doubled haploid females were first produced after gynogenetic reproduction of standard females plus inhibition of first embryonic cleavage59 , and further reproduced by a second round of gynogenesis (inhibition of second meiosis)60 (link). Homozygous clonal lines were further maintained during every generation by single within-line pair mating between one female and one hormonally sex-reversed male. Tissues (liver, brain, heart, skin, ovary, white and red muscle, anterior and posterior kidney, pituitary gland, stomach, gills) were collected and stored in liquid nitrogen until RNA extraction. Total RNA was extracted using Tri-reagent (Sigma, St-Louis, USA) at a ratio of 100 mg of tissue per ml of reagent according to the manufacturer’s instructions. RNA-Seq Illumina Libraries were prepared (Supplementary Methods) and sequenced using 101 base-lengths read chemistry on an Illumina GAIIx sequencer (Illumina, USA). In order to compare the expression levels of ohnologous genes, we restricted the analysis to the parts of the coding regions that can be confidently aligned using MUSCLE56 (link) between the two genes, as non-alignable or low-quality alignment regions may result from errors in the automatic annotation process. We retained regions of the alignment where the majority of codons contain at most 1 nucleotide change, and masked all other codons with Ns. We mapped RNA-seq reads to these alignable regions using BWA61 (link) with stringent mapping parameters (maximum number of mismatches allowed –aln 2). Mapped reads were counted using SAMtools62 (link), with a minimum alignment quality value (–q 30) to discard ambiguous mapping reads. The numbers of mapped reads were then normalized for each gene across all tissues using DESeq63 (link). As the alignable regions of both ohnologues are of the same length by construction, no additional normalization for length was necessary to compare expression levels within each ohnologue pair. Correlations between the expression levels of ohnologues were performed using Pearson’s correlation and paired Student’s t-tests in R on log2-transformed data. Log2-transformed expression profiles of rainbow Ss4R ohnologues were also analysed using supervised clustering methods. Hierarchical clustering was processed using centroid linkage clustering with Pearson’s uncentred correlation as similarity metric on data that were normalized and median-centred using the Cluster program64 (link). Expression levels were normalized and centred independently for each Ss4R ohnologue pair to compare expression profiles (Fig. 4a) and normalized and centred across both ohnologues to highlight differences in relative levels of expression between both ohnologous genes (Fig. 4b). Results (colorized matrix) of hierarchical clustering analyses were visualized using the Java TreeView program65 (link).