Human immunodeficiency virus (HIV) is a retrovirus that weakens the immune system and permits opportunistic diseases such as hepatitis C (HCV) to enter the body. These diseases induce metabolic disorders in the patients and it is therefore logical to approach them from a holistic, functional perspective, studying the metabolome comprehensively to identify metabolic signatures associated with certain disease states. The metabolomics strategy here proposed involves metabolic fingerprinting using Fourier transform infrared spectroscopy and chemometric tools on 72 plasma samples (subdivided into 63 training and 9 test samples) to differentiate between healthy subjects and patients with different disease stages. Several options, relating to the variable selection method used in linear discriminant analysis and the number of categories being considered, were explored to optimize discrimination ability. A total of 18 bands enabled differentiation between control subjects, HIV patients and the group that encompassed patients with acquired immune deficiency syndrome (AIDS), AIDS/HCV and HIV/HCV, providing overall classification and internal prediction rates of 97.67% and 93.65%, respectively. Only 9 bands were required to further discriminate between AIDS, AIDS/HCV and HIV/HCV, with 99.20% (training) and 89.66% (cross-validation) correct classifications. The simplicity and effectiveness of the classification methodology proposed was reinforced by the satisfactory results obtained in external prediction.