Cheminformatics approaches for metabolomics research
Metabolomics is part of the modern life sciences tree including genomics, transcriptomics and proteomics. One of the major goals in metabolomics is to identify small molecules including lipids, organic acids, sugars and amino acid and to obtain spatial and temporal snapshots of metabolite concentrations from complex biological matrices. The results are used to obtain greater insight into plant and animal metabolism and to understand complex cellular processes. Cheminformatics approaches play a crucial role in metabolomics because properties of organic molecules can be modeled in-silico and are matched against experimentally obtained result from analytical techniques including liquid chromatography-mass spectrometry (LC-MS) and gas chromatography-mass spectrometry (GC-MS). The accuracy of such predictions including octanol water partition coefficients (logP), acid dissociation constants (pKa) and boiling points (bp) are important to reduce the search space from millions of possible structural isomers down to a low number of potential candidate structures. Besides that, cheminformatics software tools must be able to handle millions of structures and a multitude of structure formats and should provide simple-to-use graphical front ends for novice users and versatile command line parameters for experts. We used Marvin GUI components including ChemAxon MarvinView, Standardizer and Reactor as well as the JAVA API components in a metabolomics environment to generate molecular structures and curate mass spectral databases. The use of such databases allows the measurement of more than 150 metabolites in a single sample, requiring only few micrograms of biological material. Using our high-throughput analytical techniques we are able to profile several thousand samples per year. In addition to that, we show how the Instant JChem program with its integrated overlap search and Marvin tools can be used as a general purpose tool for structure handling. We show how physico-chemical property calculations obtained from Marvin are used to support the structure elucidation process in metabolomics.