The importance of poly-unsaturated fatty acids (PUFAs) in food is crucial for the animal and human development and health. As a complementary strategy to nutrition approaches, genetic selection has been suggested to improve fatty acids (FAs) composition in farmed fish. Gas chromatography (GC) is used as a reference method for the quantification of FAs; nevertheless, the high cost prevents large scale phenotyping as needed in breeding programs. Therefore, a calibration by means of Raman scattering spectrometry has been established in order to predict FA composition in rainbow trout Onchorhynchus mykiss adipose tissue. FA composition of visceral adipose tissue was analysed by both GC and Raman micro-spectrometry techniques on 268 individuals fed with three different feeds, which have different FA compositions. Among the possible regression methods, the ridge regression method, was found to be efficient to establish calibration models from the GC and spectral data. The best cross-validated R2 values were obtained for total PUFAs, omega-6 (Ω-6) and omega-3 (Ω-3) PUFA (0.79, 0.83 and 0.66, respectively). For individual Ω-3 PUFAs, α-linolenic acid (ALA, C18:3), eicosapentaenoic acid (EPA, C20:5) and docosahexenoic acid (DHA, C22:6) were found to have the best R2 values (0.82, 0.76 and 0.81, respectively). This study demonstrates that Raman spectroscopy could be used to obtain good correlation coefficients on adipocytes allowing to predict PUFAs, and calibration models can be used to predict PUFAs contents for large scale and high throughput phenotyping in rainbow trout.