For gene expression calculations (level and divergence), we first normalized the number of reads aligned per gene per individual by the median expression of each library. Normalization via the common approach of Reads per Kilobase per Million Mapped Reads (RPKM) can be strongly biased by a relatively small proportion of highly expressed gene and we therefore decided against using it . Given that we were also interested in interpreting differences in expression between genes, we normalized all expression data by gene length. Then, we used the r package deseq to calculate expression divergence per gene, as the absolute mean value of expression for species 1 divided by expression for species 2 on a base two logarithmic scale.