DATA MINING PATIENT RECORDS TO DISCOVER DISEASE LINKS QUICKLY

Posted on August 31, 2011 by Jeffrey Newman

With any treatment of chronic disease, speed and precision are both essential. Danish scientists have created a new integrated data mining tool that extracts information from clinician’s notes and genetic info that can reveal links between health problems in a patient that would otherwise seem unrelated. The research report was published on August 25 in PloS Computational Biology. The findings found what are considered relevant correlations, for example, between migraines and hair loss and glaucoma and hunching back. In addition to generating new leads on the molecular working of disease and potential connections between diseases including neurological disorders, the model is revelaing a richer portrait of each patient. The study co-author is Soren Brunak of the Center for Biological Sequence Analaysis at the Technical University of Denmark in Lyngby and the University of Copenhagen. Using health codes for classifying diseases the researchers created a map linking more than 4,700 patients at a large psychiatric hospital by diagnosis. They integrated that data with 10 years worth of information via data mining from clinicians’ notes, an average of 25,000 words per patient. More than 800 pairs of health problems showed up and 93 were flagged by doctors as potentially significant. The team identified a connection in proteins between hair loss and migraines. Although no cures were found in this broad study, the importance is the methods being used to help researchers and physicans to find more promising leads to disease treatments instead of the arcane and lethargic scientific methods presently in use. The information will be particularly valuable to researchers laboring to find treatments for complex diseases like alzheimer’s and Parkinson’s.  False positives can be excluded based on researchers’ knowledge of cellular mechanisms and known cellular functions. This kind of threshold sifting is relevant to staying on the real leads and not getting lost in false correlations simply because two things appear to be connected. The data analysis has a primary benefit in  helping researchers in the lab to see connections they haven’t thought of before but which are directly related to the work they are doing. New and similar approaches are being developed to examine medical records in medical facilities in this country. The hope and probability is that this approach will speed knowledge acquisition concerning linkages between diseases and health disorders not previously known. It will enable researchers to think laterally in ways they not otherwise have considered. This could also be the foundation of an evolution in disease understanding and broader knowledge of cellular mechanisms as a result. This basic advance in human knowledge resulting from the good leads tapped through data mining, which is fast in relative terms, should help break through the present stalemate in finding therapeutics to the more difficult chronic diseases.