The identification of molecular biomarkers in CSF from individuals affected by Huntington disease may help improve prediction of disease onset, better define disease progression, and could facilitate evaluation of potential therapies. The primary objective of our study was to assess previously reported changes in CSF protein markers and investigate novel exploratory protein candidates in CSF from Huntington disease mutation carriers and healthy controls. Our secondary objective was to compare the discriminatory potential of individual protein analytes and combinations of CSF protein markers to assess their sensitivity and specificity in distinguishing between disease stages. We conducted a hypothesis-driven, parallel-reaction monitoring mass spectrometry-based analysis of 26 pre-specified protein analytes in CSF from sixteen manifest Huntington disease subjects, eight premanifest Huntington disease mutation carriers and eight healthy control individuals. In addition to reproducing reported changes in previously investigated CSF biomarkers (NEFL, PDYN, and PENK), we also identified novel exploratory CSF protein analytes whose levels were altered in Huntington disease mutation carriers and/or across disease stages. We also observed significant associations of select CSF protein analytes with clinical outcomes of disease severity. Using sparse partial least square discrimination analysis, we showed that a combination of CSF protein markers can distinguish between premanifest and manifest Huntington disease individuals with high discriminatory accuracy. Finally, we performed an exploratory combinatorial receiver operating characteristic curve analysis and identified optimized panels of specific CSF protein markers that can discriminate individuals with high sensitivity and specificity based on Huntington disease mutation status and disease severity. In this study, we demonstrate that combinations of CSF proteins can outperform individual markers for classifying individuals across disease stages. Moreover, we define novel combinations of CSF proteins that may be useful for improving prediction accuracy of disease onset and complement existing clinical and imaging biomarkers for monitoring Huntington disease severity.