Under global climate change, the Qinghai-Tibet Plateau (QTP) has experienced dramatic environmental changes, including salinity changes of water bodies, which may threaten the biodiversity of aquatic organisms on the ‘roof of the world’. The Tibetan loaches (the genus Triplophysa) are the largest component of the QTP ichthyofauna. Here we compared the population structure and adaptive mechanisms to salinity of two sympatric Tibetan loach species, T. stewarti and T. stenura, using population genomics methods. Using both the ‘common’ and ‘individual’ SNP datasets of seven populations from five localities of the two species, we found out that the two species showed entirely different patterns of population differentiation, with T. stewarti populations deeply diverged and T. stenura populations mixed. We identified a catalogue of candidate genes possibly involved in salinity acclimatation of the two species using both unsupervised and supervised population differentiation methods. In this section, a new approach - linkage disequilibrium (LD) graph learning - was developed and utilized to identify clusters of loci showing similar genetic differentiation patterns. However, we found limited parallel adaptive signals to salinity of the two species using these methods. Our findings broaden our understandings of the population characteristics and adaptive mechanisms of these previously underexplored Tibetan loach species, and will play a role in the biodiversity protection of Triplophysa species on the QTP.

The files include SNPs and quantitative expression data based on RNA-seq underlying this study, as well as code for the newly developed unsupervised linkage disequilibrium (LD) graph learning method.

'two_species_common_snp_dataset.vcf.gz' - the common SNP dataset for the two species, Triplophya stewarti and T. stenura.

'Tstewarti_individual_snp_dataset.vcf.gz' - the individual SNP dataset for T. stewarti.

'Tstenura_individual_snp_dataset.vcf.gz' - the individual SNP dataset for T. stenura.

'expression_gene_count_matrix.csv' - the expression gene count matrix based on RNA-seq of the two species.

'LD graph learning.R' - code for linkage disequilibrium (LD) graph learning method.