Fingermark patterns are one of the oldest means of biometric identification. During this last decade, the molecules that constitute the fingermark residue have gained interest among the forensic research community to gain additional intelligence regarding its donor profile including its gender, age, lifestyle or even its pathological state. In this work, the molecular composition of fingermarks have been studied to monitor the variability between donors and to explore its capacity to differentiate individuals using supervised multi-class classification models. Over one year, fingermarks from thirteen donors have been analysed by Matrix-Assisted Laser Desorption/Ionisation Mass Spectrometry Imaging (n = 716) and mined by different machine learning approaches. We demonstrate the potential of the fingermark chemical composition to help differentiating individuals with an accuracy between 80% and 96% depending on the period of sample collection for each donor and size of the pool of donors. It would be premature at this stage to transpose the results of this research to real cases, however the conclusions of this study can provide a better understanding of the variations of the chemical composition of the fingermark residue in between individuals over long periods and help clarifying the notion of donorship.