LEADER 00000nam  2200349   4500 
001    AAI3173947 
005    20061206120143.5 
008    061206s2005                        eng d 
020    9780542113949 
035    (UnM)AAI3173947 
040    UnM|cUnM 
100 1  Kharchenko, Peter Vasili 
245 10 Local properties of metabolic networks 
300    140 p 
500    Source: Dissertation Abstracts International, Volume: 66-
       05, Section: B, page: 2393 
500    Adviser:  George M. Church 
502    Thesis (Ph.D.)--Harvard University, 2005 
520    Metabolism encompasses essential processes of living 
       organisms. Reconstruction of detailed metabolic models of 
       several organisms provides the means to analyze functional
       and organizational aspects of metabolism on a system-wide 
       scale. The presented work investigates how functional 
       requirements of metabolism manifest themselves in 
       topological organization of the metabolic network 
520    Our analysis begins by studying patterns of co-regulated 
       genes in a metabolic network. Based on mRNA co-expression 
       patterns and distribution of known transcription factor 
       binding sites, we illustrate predominance of local 
       regulation of the metabolic network. We show that 
       strongest co-regulation is observed in specific network 
       topologies of metabolically adjacent enzymes. At higher 
       levels, we demonstrate localized co-regulated regions 
       spanning increasing portions of the metabolic network. The
       second chapter illustrates that local topological 
       properties of the metabolic network, such as connectivity,
       correlate with evolutionary and functional characteristics
       of the metabolic enzymes. Furthermore, we show that local 
       property can be observed in many types of functional and 
       evolutionary association, such as co-occurrence in 
       phylogenetic profiles or clustering of genes on the 
520    Based on these observations we develop a method for 
       identifying candidate genes encoding a specific metabolic 
       function. We illustrate that the approach is applicable to
       many types of associating evidence, and develop efficient 
       ways of combining such information. We demonstrate that 
       the method is able to predict approximately 60% of known 
       enzyme-encoding genes within the top ten candidates for 
       their enzymatic function. The predictions generated by 
       this method are complementary to traditional homology-
       based identification approaches 
590    School code: 0084 
590    DDC 
650  4 Biology, Genetics 
650  4 Biophysics, General 
690    0369 
690    0786 
710 20 Harvard University 
773 0  |tDissertation Abstracts International|g66-05B 
856 40 |uhttp://pqdd.sinica.edu.tw/twdaoapp/servlet/