We performed metabolic profiling on yellowtail (Seriola quinqueradiata) muscle to develop an objective taste evaluation method for fish meat. Dark (DM) and ordinary (OM) muscle samples before and after storage were subjected to gas chromatography-mass spectrometry (GC-MS) analysis and taste measurements using an electronic tongue. The metabolites identified by the GC-MS analysis were treated as x variables, and the taste values obtained by the electronic tongue were treated as y variables. The relationships between the metabolites and taste attributes were evaluated by two-way orthogonal projections to latent structures (O2PLS) analysis. The O2PLS analyses were normalized in two ways, unit variance (UV) and pareto (Par) scaling. The O2PLS (UV) analysis produced 3+1+0 models in Autofit and this model was statistically significant with R2Y (0.73) and Q2 (0.52) metrics. In particular, significant correlations were found between DM or OM and metabolite intensity and taste attributes, and strong associations were found between “sourness” and lysine, “irritant” and alanine and phenylalanine, “saltiness” and pantothenic acid, and “umami” and creatinine and histidine. The O2PLS (Par) analysis of DM generated significant predictive models for “acidic bitterness,” “irritant,” “saltiness,” “bitterness,” “astringency,” and “richness.” Among these, only “irritant” was affected by storage. This method was thus effective in evaluating the taste of yellowtail muscle.