Partial Least Squares (PLS) regression was used to predict the sensory characteristics (odour and taste) of industrial and artisanal Manchego cheeses, as expressed by physico-chemical parameters, proteolysis variables and certain organic acids. PLS regression demonstrated that the variables most contributing to prediction of the olfactory profile were: pH, total nitrogen (TN), αs2 + β-CN, γ-CN, citric acid and acetic acid; these variables, plus the water-soluble nitrogen fraction (WSN), contributed most to predicting the taste profile. The models obtained yielded good results for the prediction of sensory characteristics of Manchego cheeses: the root mean square error of calibration (RMSEC) and the root mean square error of prediction by cross-validation (RMSECPV) were below 1.1 for the unstructured 10-point scale used in descriptive sensory analysis. These models may thus provide a useful tool to describe the sensory characteristics of Manchego cheeses.