Metabolic pathway prediction based on inclusive relation between cyclic substructures

Abstract

Secondary metabolites of the plant kingdom have long been important as leading precursors in the pharmaceutical industry (Simmond and Grayer 1999). Reconstruction of biopathways in plants plays key roles in effectively biosynthesizing those precursors, but rational engineering of secondary metabolic pathways in plants requires a thorough knowledge of the whole biosynthetic pathway and a detailed understanding of the regulatory mechanisms controlling the onset and flux of the pathways. Such information is not yet available for the vast majority of secondary metabolites though studies have progressed extensively. For example, chemical structures of around 50,000 secondary metabolites from the plant kingdom have been determined (Verpoote 1998;De Luca and St Pierre 2000) whereas only around 2,000 enzyme reactions are known. Some researchers have predicted more than 200,000 metabolites for the plant kingdom and the number of plant species is predicted to be around 400,000 in the world (Hostettmann 2000). Thus, experimental evidence is insufficient for assigning all metabolites to metabolic pathways.There are several approaches for pathway prediction such as fingerprints (Tohsato and Nishimura 2008), reaction rule-base (Langowski and Long 2002;Talafous et al. 1994;Ellis et al. 2006;Hou et al. 2004;Oh et al. 2007) and maximum common subgraph search (MCSS) (Kotera et al. 2008). The fingerprint-based approach predefines some important molecular fragments and determines which fragments are included in each metabolite as bit-strings consisting of 0's and 1's. This approach can measure similarity between two metabolites without much computational effort, but these fingerprints can't consider connectivity between each fragment, making it difficult for this approach to predict correct pathways. Rule-based approaches predefine reaction rule-base on the basis of predefined organic metabolic reactions and predict possible pathways. Prediction by these approaches has the limitation that it depends on the restricted types of rule-base. Also, there is the possibility that the result of prediction of unknown pathways is biased by the nature of known pathways. An MCSS-based approach does not require predefined information, but this approach is an NP-hard problem (Hattori et al. 2003). Therefore, an MCSS-based approach is computationally difficult and probably to reduce the burden of computation in one such approach, only 2,502,333 metabolite pairs were compared while

Supervised de novo reconstruction of metabolic pathways from metabolome-scale compound sets

“…The computational time of the proposed method was remarkably smaller than that of the rule-based method (Nakamura et al , 2012) that needed 0.03 s per pair (corresponding to ∼2000 h for all possible compound pairs). The same task is not feasible by other reaction-filling framework methods because of their methodological limitation (Kotera et al , 2008; Tanaka et al , 2009). The compound–filling framework methods (Gao et al , 2011; Moriya et al , 2010) needed tens of seconds per compound (corresponding to thousands of hours for all possible compound pairs).…”

“…Some of the previous methods depend on predefined chemical transformation patterns (Hatzimanikatis et al , 2005; Nakamura et al , 2012). The other previous methods reduce this dependency by comparing chemical graph structures of all compounds in databases and by determining possible chemical transformations (Kotera et al , 2008; Tanaka et al , 2009), but they suffer from huge computational costs. Thus, large-scale prediction is not computationally feasible (Kotera et al , 2008), or its applicability is limited only to ring-structured compounds (Tanaka et al , 2009).…”

Metabolic pathway reconstruction strategies for central metabolism and natural product biosynthesis

“…The reaction-filling methods can be classified into those that depend on pre-defined chemical transformation rules [ 62 , 63 ] and those that do not [ 64 , 65 ]. These methods face what can be regarded as a problem of enzymatic reaction-likeness, i.e.…”

Metabolome-scale de novo pathway reconstruction using regioisomer-sensitive graph alignments

“…The overall problem resembles that of synthetic organic chemistry ( Faulon and Sault, 2001 ), but few studies have tackled this problem for enzymatic reactions. Previously developed de novo methods can be categorized into either the compound-filling framework ( Darvas, 1988 ; Ellis et al , 2008 ; Greene et al , 1999 ; Moriya et al , 2010 ; Talafous et al , 1994 ) or the reaction-filling framework ( Hatzimanikatis et al , 2005 ; Nakamura et al , 2012 ; Tanaka et al , 2009 ). However, previous methods in both frameworks are not applicable to metabolome-scale compound sets because of prohibitive computational burden.…”