An Algorithm for Distinguishing an Individual's Disease

Our hypothesis is that diagnosis of an individual's type of disease, such as influenza or cancer, can be distinguished without the need for a "fingerprint" library, which previous integrated methods have required. Typical methods which integrate a variety of data have required comparison of other diseased tissues, i.e. the need to build up a library of diseased tissue types.  We have eliminated the need for such a fingerprint library: We require a variety of data for diagnosis, however our method is not based on a library but rather on a measure which integrates an individual's healthy versus diseased data. We tested our methodology on the H1N1 influenza A of 2010-2011 United States hemagglutinin gene sequences. We were able to distinguish human host H1N1 influenza A and swine host H1N1 influenza A. We found a statistically significant difference between the human and swine hosts, as well as a measure for distinguishing the hosts. We employed both numerical linear algebra and the Procrustes' algorithm to achieve our aims.