Biomarker discovery in cerebrospinal fluid (CSF) can provide important information helping to diagnose and monitor disease progression in brain disorders. We present a label-free LC-MS strategy to investigate the proteomic profile of CSF. We provide a framework and protocol addressing quality control, sample replication steps and the adaptation of pattern recognition methods for the detection of experimental variation and (most importantly) putative biomarkers. This strategy was tested using a total of 20 CSF samples (ten samples from healthy volunteers and ten from schizophrenia patients). A clear difference was found between healthy volunteers and schizophrenia patients. With technical and biological variation controlled, we were able to detect 77 proteins with confidence in CSF, of which seven were newly identified, without prior depletion of abundant protein species. Our strategy provides a simple and controlled method for CSF biomarker discovery.