A Semantically Enabled Framework for Small Molecule Metabolic Fate Prediction: the Web as a Biochemical Reactor

Prediction and dynamical characterization of metabolic fate of small molecules is essential to understanding chemical toxicity, activity, and mechanisms of bioaccumulation, degradation, and industrial "green" synthesis. Unfortunately, experimental methods for metabolic fate characterization are prone to missing transient metabolites, experience difficulty in adjusting to variations in metabolic networks, and are expensive in the face of the enormous chemical space that requires characterization. The quantitative structure-activity relationship-based methods employed currently to predict metabolic fate are insufficient since they often ignore the dynamics of metabolism or are limited in scope and applicability. First principle-based approaches currently require expensive manual coordination of tools and formalisms to develop a cohesive dynamical model. An integrative framework to support these efforts is vital for metabolomics research to progress efficiently while taking full advantage of the expanding biochemical information and improving computational resources.;Although the framework developed here does not aim to address all aspects of metabolism prediction, or to cover any appreciable portion of all human enzymes, it demonstrates the feasibility of transforming the Web into a unified, reconfigurable, distributed biochemical reactor and can open new possibilities for interdisciplinary research.;In this work, I investigate the feasibility of constructing such a framework to integrate the computational and informational resources required for automated metabolic fate prediction of small molecules. To this end, I demonstrate a framework, based on semantic web technologies, which attempts to connect the highly fragmented world of chemical information at the levels of data representation and computational resource interoperability. This modular prototype framework is reconfigurable and adjustable for future advancements in computational chemistry and improvements in computational power. Within this framework, enzymes are distributed as semantically annotated web services, each with logically defined molecular input and output classes. A machine client that is capable of automatically classifying small molecules strings together biochemical pathways by calling upon the appropriate distributed enzyme services and obtaining the kinetic parameters and reaction specifications for each reaction. Kinetic reaction analysis is then carried out on the resultant pathway and the toxicity of each chemical entity is assessed using semantically encoded toxicity decision trees.