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

Abstract

Motivation: Recent advances in mass spectrometry and related metabolomics technologies have enabled the rapid and comprehensive analysis of numerous metabolites. However, biosynthetic and biodegradation pathways are only known for a small portion of metabolites, with most metabolic pathways remaining uncharacterized.Results: In this study, we developed a novel method for supervised de novo metabolic pathway reconstruction with an improved graph alignment-based approach in the reaction-filling framework. We proposed a novel chemical graph alignment algorithm, which we called PACHA (Pairwise Chemical Aligner), to detect the regioisomer-sensitive connectivities between the aligned substructures of two compounds. Unlike other existing graph alignment methods, PACHA can efficiently detect only one common subgraph between two compounds. Our results show that the proposed method outperforms previous descriptor-based methods or existing graph alignment-based methods in the enzymatic reaction-likeness prediction for isomer-enriched reactions. It is also useful for reaction annotation that assigns potential reaction characteristics such as EC (Enzyme Commission) numbers and PIERO (Enzymatic Reaction Ontology for Partial Information) terms to substrate–product pairs. Finally, we conducted a comprehensive enzymatic reaction-likeness prediction for all possible uncharacterized compound pairs, suggesting potential metabolic pathways for newly predicted substrate–product pairs.Contact: maskot@bio.titech.ac.jp

Simultaneous prediction of enzyme orthologs from chemical transformation patterns for de novo metabolic pathway reconstruction

“…The design of chemical transformation patterns of substrate–product pairs is crucial for the task of enzyme prediction. We represented each substrate–product pair by a high-dimensional descriptor based on chemical substructure changes between a substrate and a product using Pairwise Chemical Aligner (PACHA), because it worked the best among existing chemical descriptors for enzymatic reaction-likeness prediction according to previous work ( Yamanishi et al , 2015 ). We applied PACHA to perform a chemical graph alignment in order to detect chemical changes between two chemical compounds, and represented each substrate–product pair as an integer-valued vector (the PACHA feature vector) that describes conserved atoms, as well as generated and eliminated bonds.…”

“…The de novo reconstruction of metabolic pathways has two goals: (i) elucidation of putative reactions (previously unknown reactions) among compounds and (ii) elucidation of the associated enzymes catalyzing the putative reactions. Toward the first goal, several in silico methods have been developed based on the chemical structures of compounds by hypothesizing intermediate compounds necessary between the source and target compounds ( Darvas, 1988 ; Ellis et al , 2008 ; Faulon and Sault, 2001 ; Greene et al , 1999 ; Moriya et al , 2010 ; Talafous et al , 1994 ) or by predicting the enzymatic reaction-likeness among many compounds; that is, whether given pairs of compounds can be chemically interconverted by single enzymatic reactions ( Hatzimanikatis et al , 2005 ; Kotera et al, 2008 , 2013 , 2014a ; Nakamura et al , 2012 ; Yamanishi et al , 2015 ). All of these previous methods fail to address the second goal, i.e.…”

“…This approach enables the prediction of putative reaction networks among compounds and the prediction of the associated enzyme orthologs catalyzing the putative reactions in a seamless manner. We have already established an efficient supervised method that enables the prediction of putative reaction networks among compounds ( Kotera et al , 2013 , 2014a ; Yamanishi et al , 2015 ). In this work, we develop a novel algorithm to predict enzyme orthologs catalyzing the putative reactions.…”

Metabolic pathway reconstruction strategies for central metabolism and natural product biosynthesis

“…Reactions in metabolic pathways do not occur at random: reaction sequences exhibit many conserved patterns, termed reaction modules [ 70 , 71 ]. The concept of “enzymatic reaction-likeness” can be generalized as “ k -step reaction sequence-likeness”, predicting how many ( k ) reactions are required to convert the starting compound into the goal compound [ 72 ]. Intermediate compounds are also predicted by a recursive procedure using step-specific classifiers (that predict the n -th compounds in the k -step reactions) based on chemical substructures.…”

Dual graph convolutional neural network for predicting chemical networks

“…Statistical machine learning methods such as kernel methods have also been successfully applied to chemical property prediction [19][20][21]. In addition, statistical machine learning methods have been applied to predicting chemical networks such as metabolic reactions [22][23][24][25][26][27], drug-drug interactions [28][29][30][31] and beneficial drug combinations [32,33] by taking a pair of compounds as an input to a classifier.…”