Managing the welfare of fish in captivity is crucial to ensure a sustainable aquaculture production. Physiological stress responses, such as the plasma levels of cortisol, glucose and lactate, are the most common indicators of assessing farmed fish welfare, but their reliability has been questioned due to high biological variability and fish adaptation processes. An integrated multi-omics approach can be a promisor strategy to discover more robust fish welfare biomarkers, since it can offer the possibility of understanding the complete flow of information in the fish biological system. The aim of this work is to use proteomics to identify a restricted protein map as putative fish welfare biomarkers and integrate these results with transcriptomics and metabolomics data to achieve a global picture of the fish response to stress. Sparus aurata was reared under different stressful conditions: overcrowding, repetitive net handling (air exposure), and hypoxia, using fish reared under optimal conditions as control. Fish were sampled after 45 days of trial and proteins extracts were prepared from liver samples. Proteins were separated by 2D-DIGE and identified by MALDI-TOF/TOF MS. Putative welfare biomarkers were then chosen based on their stress-related function, fold-change and score, and used for primer design. Total RNA was extracted from liver samples using Trizol reagent with DNase treatment and used for cDNA synthesis. The mRNA levels of the target genes were assessed by real-time PCR. Comparative proteomics showed, in the liver, a total of 147 proteins statistically different in their abundance among conditions, from which 24 were indicated as putative welfare biomarkers and chosen for their transcription level analysis. Quantitative gene expression analysis reveals that the levels of transcripts of 7 of the target genes were modulated. This joint analysis provides a starter point for the development of more reliable fish welfare assessment measures to improve aquaculture sustainability.