Here we propose to create an intelligent set of randomly selected "disease structures" (peptides) to identify mainly tertiary and quaternary epitopes seen by the human immune system. The binding of immunoglobulins to peptides depends on its shape and charge, not the amino acid sequence, which is likely to allow us to also detect unknown antigenic carbohydrates and lipids. We have recently studied randomly generated peptides in silico for theoretical antigenicity based on previously identified B-cell epitopes. This was done by creating a new mathematical formula applying the "Markov" and "Yen's K-shortest path" algorithms We can thus screen and rank an "infinite" number of peptides for theoretical reactivity but have started with a subset (10-6). In the first round, we have studied 172 943 synthetic peptides for practical reasons as they fit on an array load. In preliminary experiments, we have screened human sera from patients with 'discrete and defined' neurological disorders, such as obsessive compulsive disorder or amyotrophic lateral sclerosis and identified novel and uniquely reactive sequences. We now propose to develop diagnostic assays to study these and other diseases where the causes are unknown together with experts in the field. In later steps, this should lead to an increasing understanding of the onset and prevention of the diseases.