The multivariate linear model is a powerful tool for examining associations between a single nucleotide polymorphism and multiple related phenotypes. To improve the computational efficiency of this complex correlation analysis, we spectrum-transformed multiple correlated phenotypes to mutually independent “super traits” by using the mixed model association analysis method of a single trait. The sum of these statistics was used to infer pleiotropic nucleotide genetic loci that control multiple related phenotypes. The multivariate linear mixing model of GEMMA software and the phenotypic orthogonalization method proposed in this study were used to analyze the genome-wide associations (GWASs) of three traits in turbot (Scophthalamus maximus): body length, body mass, and caudal peduncle width. The results showed that 11 QTNs were detected on chromosome 1, 5, 6, 8, 12, 16, 20, 21, and 22 by the multivariable linear mixed model method. However, our method detected a total of 14 QTNs on chromosome 1, 3, 5, 6, 8, 12, 16, 20, 21, and 22, among which nine QTNs were detected jointly by the two methods. Compared with that of GEMMA, our method exhibited higher power to detect QTNs of multi related traits, which provides a convenient and efficient strategy for multi-trait GWAS research and the genetic breeding of other aquatic species.