A major goal of clinical proteomics is the identification of protein biomarkers from mass spectral analyses of fairly easily obtainable samples such as blood serum, urine or cerebrospinal fluid from patient populations. It is hoped that such protein biomarkers can be utilized for early detection of disease and examined further for potential therapeutic use. In this paper, we present the process for successful discovery of biomarkers that are indicators of a chronic neurodegenerative disease of motor neurons, called Amyotrophic Lateral Sclerosis; from application of rule learning to the analysis of proteomic mass spectra from cerebrospinal fluid samples. We have implemented a wrapper-based rule learning framework within which the massive number of features that accumulate from mass spectral analyses of clinical samples can be evaluated by repeated invocation of a rule learner. Our framework facilitates evidence gathering as indicated in this case study, and can speed up disease-specific biomarker discovery from clinical proteomic mass spectra.