Personalized treatment is highly desirable in multiple sclerosis because it is an immensely heterogeneous disease. This heterogeneity is seen in both the disease course and the treatment responses. Currently, a combination of clinical features and imaging parameters in magnetic resonance imaging is used to classify active and non-active patients and treatment responders and non-responders. Although this classification works on a group level, individual patients often behave differently from the group. Therefore additional biomarkers are needed to provide better indicators for prognosis and treatment response. Basic and clinical research have discovered different promising targets. It is now essential to verify the utility and accuracy of these markers in large, prospectively sampled patient cohorts.