Ageing compromises skeletal muscle mass and function through poorly defined molecular aetiology. Here we have used untargeted metabolomics using UHPLC-MS to profile muscle tissue from young (n=10, 25±4y), middle aged (n=18, 50±4y) and older (n=18, 70±3y) men and women (50:50). Random Forest was used to prioritise metabolite features most informative in stratifying older age, with potential biological context examined using the prize-collecting Steiner forest algorithm embedded in the PIUMet software, to identify metabolic pathways likely perturbed in ageing. This approach was able to filter a large dataset of several thousand metabolites down to subnetworks of age important metabolites. Identified networks included the common age-associated metabolites such as androgens, (poly)amines/amino acids and lipid metabolites, in addition to some potentially novel ageing related markers such as dihydrothymine and imidazolone-5-proprionic acid. The present study reveals that this approach is a potentially useful tool to identify processes underlying human tissue ageing, and could therefore be utilised in future studies to investigate the links between age predictive metabolites and common biomarkers linked to health and disease across age.