2.9 Differential methylation analysis
Differential methylation analyses were performed using the DMRseq package (version 1.18.0, Korthauer 2017) with the following parameters: block = FALSE, min Num Region = 5, deltamax = 0.25, bp span = 1000, min In Span = 10, max Gap Smooth = 2500, smooth = TRUE). DMRseq output regions (grouped per 1000 bp) were considered as differentially methylated regions (DMRs) when the p-value was ≤ 0.05. Distribution of DMRs across genome features (promoter, gene body, exons, introns and intergenic regions) was performed using the bedtools intersect function (Quinlan, 2014) with the position of each genomic feature from the sea bass genome annotation file (Tine et al., 2014). DMRs located between two genomic features were counted as being located in both regions. The promoter position was defined 2kb upstream the Transcription Start Site (TSS). Hyper- and hypomethylated regions were determined according to the beta value from the DMRseq output file where a positive and a negative value correspond respectively to hyper- and hypomethylation in fresh water vs seawater. To visualize the chromosomal distribution of the DMRs, an horizontal bar plot was generated, showing the percentage of hyper- and hypomethylated regions on all D. labrax chromosomes. To determine the distribution of the DMRs within the genes among different exons and introns, we compared DMRs in exons 1 to 10 and last exons of genes as well as introns 1-10 and last introns. The analysis was performed on 10 exons as this is the average number of exons in D. labrax genes. Genes with only one exon were not included in the analysis. The sum of hypermethylated regions in all first exons was normalized by the total length (in bp) of all first exons, then this ratio was multiplied by 100 to get a percentage. The same method was used to determine a percentage of hypo- and hypermethylated regions in the other exons and introns 1-10, as well as last introns. This method enables to compare DMR frequency and distribution between different exons and introns and prevents bias due to the difference in length among exons and introns.