Rank and Order: Evaluating the Performance of Sockeye Salmon SNP Assays

The use of SNPs for the study of non-model organisms has primarily been limited by the cost and difficulties of discovering new markers, and consequently the number of available SNP markers has been small. With decreasing technology costs, SNP discovery projects are becoming more common, and the number of readily available SNP markers is continually growing. Previously, high-throughput assays for SNPs in non-model organisms were so few that every marker was considered valuable and used. Increasingly, many researchers are interested in developing SNP panels tailored to their research question and study system. SNP panels can be developed and optimized for laboratory performance, genotyping platform, and power to resolve population. Although there is currently no consensus on how to rank molecular markers, especially SNPs, ranking and evaluating a SNP’s value for a panel will be of increasing importance in ecology and resource management, especially as the number of high-throughput assays continues to grow. For sockeye salmon this is already the case. A limited set of 45 SNPs has already been a valuable tool for the study of this species, providing insight into their life history, migration, and harvest. However, the cultural and economic importance of this species across the Pacific Rim has increased demand for resolving power in such studies and created a need for more and higher resolution SNPs. Here we present 43 new SNP assays developed using next generation sequencing data. Additionally, we explore five different ranking methods for 115 SNP loci described in sockeye salmon, including those described here:  locus specific F_ST values, informativeness (I_n), average contribution of a locus to principal components (LC), and the locus ranking programs BELS and WHICHLOCI. Each of these ranking methods was used to create 48- and 96-SNP panels. Panels were then tested for performance using a f_ORCA simulation. All 96-SNP panels performed similarly better that the 48-SNP panels. Among the 48-SNP panels, panels created from F_ST, I_n, and LC ranks performed better then panels formed using the top ranked loci from the programs BELS and WHICHLOCI.  As more SNP assays become available the differences between ranking methods may be heightened and have a greater impact on panel performance, warranting careful exploration of locus ranking and evaluation methods.