Browsing by Subject "Pathway prediction"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Improved prediction of biodegradation pathways: visualization and performance.
    (2011-02) Gao, Junfeng
    The University of Minnesota Pathway Prediction System (UM-PPS) (http://umbbd.msi.umn.edu/predict/) is a rule-based system that predicts plausible pathways for microbial degradation of organic compounds. Its biotransformation rules are based on reactions found in the University of Minnesota Biocatalysis/Biodegradation Database (UM-BBD, http://umbbd.msi.umn.edu/) or in the scientific literature. Since the UM-PPS was created in 2002, its rule base has grown to 275 entries. The original system predicted one level of prediction at a time. It provided a limited view of prediction results and heavily relied on manual interventions. It matched the query compound with all biotransformation rules one by one, which was a time-consuming process. In 2008, the two-level visualization was first implemented to allow users to view two levels of predictions at a time. However, this visualization approach was usually not able to show the complete metabolism of a query compound, and users still needed expert knowledge to make educated choices to continue the prediction. In 2009, we started to develop a multi-level visualization and, simultaneously, work on increasing prediction speed. In 2010, the multi-level visualization was implemented to predict up to six levels of predictions at a time. Not only more products, but also common intermediates and cleavage products are displayed. Users can view prediction alternatives much more easily in a tree-like interactive graph. A multi-level prediction can be computationally intensive and requires users to wait longer than desired for the prediction results. Therefore, we used a multi-thread computing strategy that decreased the prediction run-time by half. We balanced the computing threads and pre-loaded all UM-PPS database tables to permit quick access to its data. Both of these improvements resulted in an additional 30% decrease in prediction run-time. We conducted a simulation study and used another web server to reduce the queuing interference by over 85%. Beta testers were satisfied with its visualization and performance. The above improvements lead to a smarter and faster UM-PPS that has continued its growth in the past 4 years. It now displays better graphical results and predicts biodegradation pathway in a timely manner.