Analysis of proteins from cerebrospinal fluid tests in search of biomarkers characterizing Multiple sclerosis

There are contradicting theories describing Multiple sclerosis (MS). This study attempts to understand MS through interpreting the bio-markers of MS. Recursive feature elimination with cross validation (RFECV) was used to select bio-markers (proteins) for MS and to detect inflammation in patients. Principal components plots of proteins selected to distinguish between patients with MS and patients without MS were not only successful in separating patients with MS and patients without MS, but also showed two types of MS patients. The first type was MS patients with inflammation and the second type was MS patients without inflammation. This finding proposes inflammation to be a secondary effect of MS instead of a primary effect. The proteins in the principal component plots were related to inflammation and neuron development/regeneration. The types of proteins found together with the separate groupings of MS patients strengthen the hypothesis describing MS as a consequence of a defect in neuron regeneration where inflammation can sometimes but not always occur. This differs from the ’outside in model’ that is often referred to when explaining MS which depicts the autoimmune response as the primary cause of neuro-degeneration. Finally, Logistic regression and support vector classifiers were trained to classify patients with MS and patients with inflammation. Models distinguishing MS had a score of 90 percent on the test data, while models classifying inflammation had a score of 98 percent on the test data.M-D