Research Paper Ultra-low-depth whole-genome sequencing with reference panel-based imputation illuminates sex chromosome evolution and enables growth rate and disease-resistance GWAS and genomic prediction in sablefish (Anoplopoma fimbria)

Abstract Genome-wide association studies and genome prediction (GWAS and GP) in non-model species are often limited by the lack of dense genotyping arrays or by high sequencing costs. Low-depth whole-genome sequencing (WGS) combined with genotype imputation can be a cost-effective alternative. However, it is not clear how low one can go with respect to target sample depth and reference panel size. In addition, these approaches have rarely been evaluated in populations with extensive family structure. We generated a sablefish ( Anoplopoma fimbria ) genome reference panel by sequencing 65 wild-caught individuals to a mean depth of 14x. Approximately 4.9 million SNPs were observed. SNP density varied among chromosomes and along chromosome arms and was lowest on chromosome 14, which contains the Master Sex-Determining gene, gonodal somatic cell derived factor ( gsdf ). We used this reference panel to impute genotypes in a target population of 1,380 farm-raised sablefish sequenced to a mean depth of 0.4x. Imputation accuracy was estimated by comparing imputed genotypes to those obtained from a PCR-based SNP genotyping panel and by comparing genotypes obtained from 14x alignments to those obtained from imputed down-sampled data. A total of 1.2 million SNPs (MAF ≥ 0.05) were utilized for GWAS and GP for weight at age 2.5 years (all target population individuals) and for sex and disease status (603 target population individuals sampled during an Aeromonas salmonicida outbreak). Significant associations with sex spanned chromosome 14. The lowest p -value SNP was 9 kb upstream of gsdf . Chromosome 14 SNPs were also associated with weight variation. This result was anticipated because females are known to grow faster than males. When only female data were analyzed, SNPs in macrophage mannose receptor 1 ( mrc1 ) were associated with weight variation. SNPs associated with disease status spanned a 46 Kb region of chromosome 20 with several candidate genes including peroxynitrite isomerase THAP4 ( thap4 ). Genomic prediction models were developed using SNPs with low GWAS p -values. Sex and disease status were predicted with multi-locus models and had classification accuracies of 0.95 and 0.71, respectively. The best weight-prediction model achieved a Pearson correlation coefficient of r = 0.65 between observed and predicted weights. Ultra-low-depth WGS with imputation using a diverse but small reference panel provided an enormous set of reliable markers for GWAS and GP in this non-model aquaculture species.