Gleason grade is the most important determinant of prognosis and survival in prostate cancer, and is determined using prostate biopsy. Here we investigate whether multi-parametric MRI can be used to classify Gleason grade non-invasively with logistic regression (LR) models, classifying tumours into 3+3 and those containing a 4 component. A selection of clinical and quantitative MRI metrics were used. The LR model was trained in ninety-nine patients and tested following a Leave-One-Out (LOO) analysis on a temporal separated cohort of nineteen patients. LR models were shown to predict the presence of Gleason 4 component in cancer lesions both before and after LOO analysis.