Research Projects

This section introduces current projects within the Systems Biochemistry Group. If you are interested in our research and would like to work within the group, even if no specific position is advertised, please send your CV to Ronan Fleming, with a cover letter explaining your research interests.

Optimization algorithms and software for variational kinetic modeling

In experimental systems biology, the majority of high throughput experimental data is of molecular abundance and the minority is of reaction rates. We seek a modeling framework flexible enough to integrate experimental data on both rates and abundance. Linear flux balance analysis does not explicitly represent the abundance of each molecule. Without explicitly representing abundance, the incorporation of experimental constraints from measurement of molecular abundance is always an approximation. Such approximation may be useful in the short to medium term, but ultimately we seek a computational model of biochemical reaction networks to explictily represent the abundance of each molecule and the rate of each reaction.

We are developing a novel Variational Kinetic approach to model stationary state reaction kinetics for large systems of reactions based on nonlinear continuous optimization algorithms. These variational kinetic models aim to preserve the computational tractablity associated with numerical optimization and add the biochemical realism typical of kinetic models. Building on our previous development of a globally convergent algorithm for the forward problem (computing reaction rates and molecular abundance given parameters), we aim to develop a globally convergent algorithm for the corresponding inverse problem (computing parameters given reaction rates and molecular abundance), via gradient based search of kinetic parameters that optimally fit experimental data.

Our general approach is to focus on the development of biochemically tailored polynomial-time optimization algorithms with guaranteed convergence properties. This effort requires the development and application of mathematical concepts at the intersection of Variational Analysis, Continuous Nonlinear Optimization, Generalised Convexity and Numerical Analysis. In tandem with mathematical algorithm development, we prototype numerical analysis software then test its performance when applied to a set of low, medium and high dimensional biochemically relevant modeling problems. Prototype software is developed in high-level languages, e.g., MATLAB. When mature, the software is reimplemented in low level languages, e.g., FORTRAN, tailored to run on high performance computing architectures.

This research is carried out in collaboration with Prof. Michael Patriksson, Mathematical Optimization Group, Chalmers University, Prof. Michael Saunders, Systems Optimization Laboratory, Stanford University and Prof. Bernhard Palsson, Systems Biology Research Group, UCSD.

  • 2012-17: U.S. Department of Energy & National Institutes of Health interagency, collaborative research award, "Multiscale Molecular Systems Biology: Reconstruction and Model Optimization", with Prof. Michael Saunders, Stanford Systems Optimization Laboratory and Prof. Bernhard Palsson, Systems Biology Research Group, UCSD.
  • 2009-12: U.S. Department of Energy collaborative research award, "Numerical Optimization Algorithms and Software for Systems Biology", with Prof. Michael Saunders, Stanford Systems Optimization Laboratory and Prof. Bernhard Palsson, Systems Biology Research Group, UCSD.

Reconstruction and computational modeling of dopaminergic neurons

Fig. 1: Camera lucida reconstruction of a dopaminergic neuron from the dorsal tier of the substantia nigra pars compacta of a rat. The total axonal length is 467,000 μm with very little arborization outside the striatum. A single dopamine neuron can influence upto 5% of all neurons in the neostraitum or ~75,000 neurons. A single neostraital neuron is under the influence of 100-200 dopaminergic neurons, with a similar redundancy is observed in humans . The axon fibers in the striatum (A) and dendrites (B) in the substantia nigra pars compacta substantia nigra were projected onto a parasagittal plane and superimposed from the medial side. (C) The dorsal and frontal views of the intrastriatal axonal arborization were reconstructed and compared with the medial view. Red and blue lines in the striatum indicate the axon fiberslocated in the striosome and matrix compartments, respectively. ac, Anterior commissure; cc, corpus callosum; cp, cerebral peduncle; CPu, caudate–putamen(neostriatum); Hpc, hippocampus; ic, internal capsule; LV, lateral ventricle; ml, medial lemniscus; ot, optic tract; STh, subthalamic nucleus; str, superior thalamic radiation; Th, thalamus; ZI, zona incerta (Figure adapted from: W. Matsuda, T. Furuta, K. C. Nakamura, H. Hioki, F. Fujiyama, R. Arai, and T. Kaneko. Single nigrostriatal dopaminergic neurons form widely spread and highly dense axonal arborizations in the neostriatum. The Journal of Neuroscience, 29(2):444–53, 2009.).

 

We are reconstructing the salient biochemistry of dopaminergic neuronal metabolism as a first step toward development of a whole cell dopaminergic neuronal model. Omics data are combined with published algorithms to generate a draft dopaminergic specific reconstruction, to obtain a subset of the metabolic reactions in the recently published cell type unspecific reconstruction of human metabolism, Recon2. Extensive manual curation of literature is used to reconcile the draft reconstruction with known biochemical features of normal dopaminergic neurons. Thermochemical calculations are used to predict the direction of net flux within each metabolic reaction. Using standard procedures, the reconstruction is converted into a constraint-based computational model of dopaminergic metabolism. Iterative rounds of reconstruction, model prediction and reconciliation with existing experimental data is used to develop a computational model which is formal synthesis of current knowledge on dopaminergic neuronal metabolic function. We test the accuracy of the dopaminergic neuronal metabolic model by comparing predicted active pathways with pathways established experimentally to be active. The computational model of dopaminergic neuronal metabolism is used as an aid to interpret experimental data, to optimally design in vitro experiments with dopaminergic neurons, to understand the aetiopathogenesis of Parkinson’s disease and to develop new approaches for early diagnosis and treatment of Parkinson’s disease.

This work is in collaboration with Prof. Ines Thiele, Molecular Systems Biomedicine Group, LCSB and Prof. Rudi Balling, Experimental Neurobiology Group, LCSB.

Automated microfluidic cell culture and imaging of dopaminergic neurons

In order to create an accurate computational model of dopaminergic neuronal metabolism, quantitative data is required to set the boundary conditions for the corresponding model. Microfluidic cell culture chips can be designed with various advantages over macroscopic cell culture, including increased precision and automation of various experimental conditions, ease of conducting perfusion culture, direct coupling to downstream analysis systems and the potential for real time on-chip analyses. High throughput experiments can also be cost efficient due to reduced reagent consumption.

We have established a versatile microfluidic cell culture platform at the LCSB. The platform (Figure 1a) is based on a PDMS microfluidic chip fabricated by the Stanford Microfluidics Foundry for 2D culture (Figure 1b) . In addition, a microfluidic titer plate developed by Mimetas (Figure 2a), based on the use of phase guide technology (Figure 2b), can be used for stratified 3D cell culture using hydrogels. The chip for 2D culture and the microfluidic titer plate for 3D culture can be mounted on a motorized stage of a motorized fluorescent microscope (Leica DMI6000B), equipped with an incubation system for live-cell imaging and a sCMOS camera (Neo 5.5, Andor Technology) for automatic recording of cell density, viability, morphology using immunostaining assays and live imaging experiments, particularly calcium imaging (Figure 2c). Everything within the microfluidic platform is fully programmable and computer operated via MATLAB. By mixing fluid inputs we are able to use the microfluidic systems to provide fresh media that more closely matches plasma concentration, offering the capacity to modulate each metabolite concentration between physiologically realistic bounds. We are exploring microfluidic cell culture of a variety of in vitro models of dopaminergic neurons, including transformed cells, primary dopaminergic neurons and dopaminergic neurons derived from induced pluripotent stem cells (Movie 1).

This work is carried out in collaboration with Dr. Rafael Gómez-Sjöberg (LBNL Microfluidics Laboratory), Dr. Paul Vulto (Mimetas), Prof. Jens Schwamborn, Developmental and Cellular Biology Group (LCSB) and Prof. Rudi Balling, Experimental Neurobiology Group, (LCSB). 

Figure 2: Microfluidic titer plate based on phase guide technology : a) top and bottom views of OrganoPlate® (b) Scheme of a single 2-lane bioreactor composed of a gel inlet (1), a perfusion inlet (2), an optical readout window (3) and a perfusion outlet (4). (c) Bioreactor in OrganoPlate® showing calcium imaging and immunostaining readouts