Transcript Abundance, Gene-Level Estimates, and Differential Expression Analysis

We quantified transcript abundance using two complementary approaches. Similar to the approach outlined above, using Salmon, we quasialigned raw reads against cDNA sequences for coding and non-coding genes available for S. trutta from Ensembl. We calculated gene-level counts using tximport and loaded these values into a DESeq2 object using DESeq2. Raw gene-level counts are provided in Supplemental Information Table 9 . We performed a Wald test implemented by DESeq2 to identify significantly differentially expressed genes between males and females (Benjamini-Hochberg (BH) adjusted p < 0.05; Supplemental Information Table 10 ). To explore similarities in expression profiles across all samples, a principal component analysis was performed with DESeq2 for all samples using gene-level counts for 33,228 genes expressed in the liver following a variance stabilization transformation implemented by DESeq2. As a complementary measure, for each individual, we examined gene expression using a traditional aligner-based approach. We aligned reads against the brown trout reference genome assembly using STAR and extracted gene-level counts from the resultant BAM files using HTSeq [v. 0.11.2; (88); Supplemental Information Table 11 ]. We found an extremely high correlation (Pearson’s correlation: r=0.9, p < 2.2e^-16) between the mean gene-level counts for both this approach, as well as the Salmon approach outlined above. This is also true for comparisons for individual samples across both approaches (lowest r=0.83, highest r=0.97). As an additional measure, we analyzed and compared the number of differentially expressed genes identified with DESeq2 when using the output of either transcript quantification approach ( Supplemental Information Table 12 ). We found that >85% of genes identified as significantly differentially expressed between the sexes by Salmon were also identified as significant by the STAR-HTSeq approach. As quasi- and pseudoaligners, such as Salmon and kallisto, have higher accuracy and consistency in transcript quantification compared to traditional aligners (89), we report only the findings of our Salmon analysis.