The efficient function of the metabolic network in eukaryotic cells relies on its separation into a number of closely-linked but metabolically distinct compartments. Perturbations in any one of these compartments alter the metabolic network throughout the cell thereby driving or contributing to the pathogenesis of many acute and chronic diseases. Determination of differential metabolic changes due to disease is very challenging because of the complexity and compartmentalisation of mammalian metabolism. Stable-isotope tracers can help to elucidate details of metabolic processes with unprecedented detail. While mass spectrometry (MS) delivers highly accurate isotopologue information even at low metabolite concentrations, the information is not very specific and sometimes ambiguous. The combination of nuclear magnetic resonance (NMR) spectra with MS derived data leads to highly resolved metabolism information for the most abundant metabolites .
The project we offer will develop novel artificial intelligence algorithms for network wide metabolic flux analysis with stable isotope tracers. One of the challenges will be to find a way to integrate genome-wide metabolic reconstruction, e.g., RECON2.2  with different experimental data such as NMR signal multiplets, MS isotopologue distributions, gene expression data and metabolite concentrations. To tackle this challenge, the student needs to develop attributed graph representation method  to integrate different experimental data. Similar to a standard graph, an attributed graph consists of nodes that connected by edges, but it also associates the nodes and edges with attributes, which can be used to store difference source of data such as metabolite concentrations (as node attributes of a metabolite node).
Based on the attributed graph representation, we will develop novel algorithms to identify active modules . These active modules are subnetworks with significantly different activities from the background network, which might be relevant to the targeted metabolic processes . In order to identify modules that strike the balance between topology and attributes, which are two independent or even conflicting source of information , we will develop a module identification algorithm based on evolutionary multiobjective optimisation algorithm . We will also investigate graph-based feature learning algorithms such as attributed graph embedding algorithms  to extract active modules from the attributed graphs.
We will also develop a completely data driven approach, i.e. the metabolic network will not be defined a priori, but the available experimental data will inform the algorithm whether certain pathways are necessary to model the metabolism or not.
This project uniquely combines a wide variety of scientific techniques, ranging from wet lab techniques such as cell culture, via the application and development of analytical techniques such as NMR spectroscopy to artificial intelligence approach such as meta-heuristic search and graph-based machine learning, which are essential for data analysis and integration of such varied data sets.
Applicants should have a strong background in physics and deep learning. They should have a commitment to perform research in an interdisciplinary environment and hold or realistically expect to obtain at least an Upper Second Class Honours Degree in physical or computational sciences.
Informal enquiries should be directed to Dr Christian Ludwig, email [email protected]
To be considered for this studentship, please send the following documents to Viktorija Ziabliceva, email [email protected]
• A detailed CV, including your nationality and country of birth;
• Names and addresses of two referees;
• A covering letter highlighting your research experience/capabilities;
• Copies of your degree certificates with transcripts;
• Evidence of your proficiency in the English language, if applicable.
 Chong et al., Angewandte Chemie, 2017, 56(15)
 Swainston, Metabolomics, 2016, 12
 Rodriguez, Bulletin of ASIS&T, 2010, 36
 Bryant, BMC Syst.Biol., 2013, 7
 Mitra et al., Nat.Rev.Genet., 2013, 14
 Zhou et al., Proc.VLDB Endow., 2009, 2
 Chen et al., BMC Syst.Biol., 2017, 11
 Goyal, arXiv:1705.0281, 2018