Updated project metadata. Here we present the data obtained from a label-free quantitative proteomics analysis of soluble spinal cord extract derived from a mouse model of multiple sclerosis (EAE) and sham-induced mice. Samples were prepared offline using the FASP approach and then submitted for nano-LC-MS/MS analysis on an Orbitrap Velos instrument. After statistical evaluation of the data, 431 differentially expressed proteins (KS-test, p < 0.05) out of a total of ~1400 unique proteins were identified in the comparative spinal cord analysis (peptide FDR=0.55%).Database search and protein identification: Tandem mass spectra were extracted from .RAW files and searched using the SEQUEST-PVM v.27 (rev.9) (Eng et al., 1994) database program against a Mouse protein database downloaded as FASTA-formatted sequences from EBI-IPI (database version 3.72) which contains 56957 entries (with priority given to UniProt identifiers) as well as reverse decoy sequences to empirically assess the false identification rate. This search program was executed on a cluster computer to match the MS/MS spectra to the corresponding most highly correlated peptide sequences Mass tolerances for precursor (MS) and product ions (MS/MS) were set to 3 and 0 m/z, respectively. Searches were performed with the enzyme selectivity set to trypsin with one missed cleavage allowed and protein modifications included fixed carbamidomethylation of cysteines (57 Da). Match likelihoods were assigned a statistical confidence score using the STATQUEST probabilistic model (Kislinger et al., 2003) and candidate peptide identifications were filtered using an estimated peptide confidence score of ≥95%. A 10 ppm high accuracy mass filter accounting for isotopic shifts in the spectra was applied post-SEQUEST analysis thus improving the fidelity of protein identifications. Protein quantitation: To estimate relative protein levels, spectral counts were transformed into normalised spectral abundance factors (NSAF) as previously described (Mosley et al., 2009). Briefly, this involves dividing the spectral count (SC) of a protein by its length (Mw) and finally normalises this value to the sum of all SC/Mw.