Meat quality (MQ) is a complex phenotype assed by different indicators measured using costly and/or invasive analyses. Early post-mortem {pm) biomarkers ofMQ have been identified, but they refer to a single MQ trait and not to the overall quality of pork samples. This study was aimed at determining pork quality classes combining both sensory and technological dimensions. Then, combinations ofbiomarkers discriminating between quality classes were identified to further predict MQ leve! of pork Joins. Sensory, technological and gene expression data were collected on 100 pig Longissimus lumborum muscle samples (last rib levet) exhibiting a wide and graduai variability in MQ. Scientific and statistical approaches were combined to select indicators and their threshold values specifying quality classes differing on sensory and technological attributes: low, acceptable, and extra quality. Gene expressions were used as predictive variables in a generalized linear mode! to discriminate quality classes. The best model included expression levels of 12 genes ( 18% error rate on known data, 24% a ft er cross validation). Based on these results, the ability to detect high quality Joins early pm could be useful in the meat industry and pork chains ofhigh quality products.