Multiple sclerosis (MS) often starts in the form of clinically isolated syndrome (CIS) and only some of the CIS patients progress to relapsing-remitting MS (RRMS). Biomarkers to predict conversion from CIS to MS are thus greatly needed for making correct treatment decisions. To identify a predictive cerebrospinal fluid (CSF) protein, we analyzed the first-attack CSF samples of CIS patients who converted (CIS-MS) (n=23) and did not convert (CIS-CIS) (n=19) to RRMS in a follow-up period of 5years using proteomics analysis by liquid chromatography tandem-mass spectrometry (LC-MS/MS) and verified by ELISA. Label-free differential proteomics analysis of CSF ensured that 637 proteins were identified and 132 of these proteins were found to be statistically significant. Further investigation with the ingenuity pathway analysis (IPA) software led to identification of three pathway networks mostly comprised proteins involved in inflammatory response, cellular growth and tissue proliferation. CSF levels of four of the most differentially expressed proteins belonging to the cellular proliferation network function, chitinase-3-like protein 1 (CHI3L1), tumor necrosis factor receptor superfamily member 21 (TNFRSF21), homeobox protein Hox-B3 (HOXB3) and iduronate 2-sulfatase (IDS), were measured by ELISA. CSF levels of HOXB3 were significantly increased in CIS-MS patients. Our results indicate that cell and tissue proliferation functions are dysregulated in MS as early as the first clinical episode. HOXB3 has emerged as a potential novel biomarker which might be used for prediction of CIS-MS conversion.