Systems Biology and Mathematical Modelling
The research focus of the group of Dr. Zoran Nikoloski is the development of methods for integration and analysis of 'omics' (time-resolved) data from high-throughput technologies in combination with the existing biological knowledge, structured in a form of ontologies and large-scale networks, in order to glean the design principles of biological processes.
The groups research focus is the development, analysis, and implementation of methods for:
- data-driven qualitative and quantitative modeling of genome-scale metabolic networks,
- evolutionary and optimisation processes in biological networks,
- automated transfer of functional knowledge within and across species,
- understanding of system behaviors/functions emerging from molecular interactions.
Methods include network optimisation, design and analysis of algorithms, mathematical programming, game theory, random processes, computational complexity.
Network-based approaches, originating from graph theory and constraint-based optimisation, lend themselves as a unifying framework for addressing the diverse modeling-related questions arising in computational-biology studies. Emphasis is placed on employing the large quantity of temporal data for modeling the dynamics of processes taking place on biological networks and the temporal changes of the networks themselves due to environmental stimuli.
Model Discrimination and Putative Allosteric Regulations
Experimental studies of photosynthesis undoubtedly point at the dependence between external conditions, inherent model characteristics (i.e., kinetic parameters), and the number of steady states. Resolving this problem would allow us to determine the conditions under which photosynthesis occurs at a steady state with higher efficiency (fixing more CO2 with a given supply of ATP) in isolated chloroplast. We work on developing methods to:
- determine the dependence of the number of steady states on the external parameters (e.g., concentrations of CO2 and phosphate),
- infer internal parameters (kinetic rates), if the model exhibits two steady states, and
- discard models from a previously established hierarchy, based on physiological feasibility of the found solutions and the compliance with experimental data.
The available data about the Calvin cycle intermediate levels can be also employed to develop data-driven methods for identifying putative allosteric regulations for model improvement. Based on the assumption that biological processes optimise a given function (e.g., yield, turnover rate, or efficiency) while demonstrating stability of the dynamic features under small perturbation, we are developing an optimisation framework for determining missing regulations in an investigated model, based on model ranking.
Environment-specific Energetic Costs of Protein Synthesis
Existing Systems Biology approaches do not provide the means for estimation and comparison of protein synthesis costs under different environmental conditions. We address questions related to different levels of amino acids and proteins observed in plants between light, dark, water deficit and different nutrient conditions and aim at providing explanations for the experimental observations with respect to the cellular economy. Our method for estimating the cost of amino acid synthesis (and, consequently, the cost of protein synthesis) is based on a genome-scale network and Flux Balance Analysis (FBA), rather than a small set of pathways, and has been used to integrate data from various cellular levels.
Elucidation of Reaction Fluxes from (labeled) Metabolomics Data
Flux Balance Analysis (FBA) has recently been extended to Dynamic FBA (DFBA), which allows for kinetic modelling and predicting changes in metabolite concentrations based solely on the system stoichiometry. However, FBA and DFBA ignore the possibility that perturbed metabolic networks may not immediately be regulated towards the optimal value of the objective. We are working on developing novel DFBA-based approaches resulting in a benchmark set of methods for simulation on networks without the use of kinetic parameters. For the case of labeled metabolomics data, we are working on extensions to the Kinetic Flux Profiling approach that can efficiently utilise the time-resolved distribution of isotopomers to resolve the flux of (ir)reversible reactions