In this paper, we illustrate how a proteomic analysis can be useful to approach complex biological problems, in this case the concept of stress-induced premature senescence (SIPS). According to the stochastic theories of ageing, damage that accumulate with time in the cellular components are responsible for cellular ageing. As a corollary, some sort of premature senescence should appear if the damage level is artificially increased due to the presence of stressing agents at subcytotoxic level. It has been shown, in several different models, that at a long-term after subcytotoxic stresses of many different natures, human diploid fibroblasts (HDFs) display biomarkers of replicative senescence (RS), which led to the concept of SIPS as compared to telomere-dependent RS. We compared RS and SIPS of HDFs by proteome analysis. SIPS was induced by two very different stressors: tert-butyhydroperoxide or ethanol. First, only a part of the protein expression changes observed in RS were also observed in SIPS. Second, HDFs in SIPS show changes specific either to the long-term effects of t-BHP or ethanol or independent of the nature of the stress. These changes have been termed "molecular scars" of subcytotoxic stresses. This work is also an excellent opportunity to discuss on important methodological issue in proteomics: the absolute requirement to start from reliable and reproducible models, which was the case in this study. We also focus on the data handling and statistical analysis allowing to use two-dimensional gel electrophoresis patterns in a semi-quantitative analysis.