Hydrodynamic and thermodynamic modeling: Using tools such as analytical ultracentrifugation and dynamic light scattering my research group focuses on the elucidation of structure and function of macromolecules, assemblies, nanoparticles and synthetic polymers. To this end, we have developed several numerical methods and computational approaches, including parallel distributed computing, for the analysis of experimental data from these hydrodynamic techniques. This long-time effort has resulted in the creation of the UltraScan software suite. By creating this software our goal is to implement novel technology in a user-friendly data analysis environment so that the methods can be applied by any investigator, even those without extensive expertise in computing, mathematics or biophysics.
A major emphasis in developing these new methods is placed on the global approach, which takes advantage of the added information content of multiple datasets from different experimental methods and experimental conditions. The global approach presents new challenges with respect to optimization algorithms and requires new paradigms to deal with the large amounts of data from combined experiments (such as multi-speed, multi-wavelength, and multi-concentration sedimentation velocity and equilibrium experiments).
Our current efforts focus on the development of novel adaptive space-time finite element solutions to partial differential equations describing sedimentation velocity experiments at a very detailed level, which extends beyond non-interacting, ideal systems to multi-component reactions, concentration dependent non-ideality, slow kinetics and reaction equilibria determinations, co-sedimenting solutes, and methods for the spectral decomposition of dissimilar absorbing components such as nucleic acids, proteins and molecules containing unique chromophores. the tcnovel analysis methods utilizing the latest advances in technology and instrumentation. Among them are parallel computational approaches using Linux Beowulf systems. Such tools are required to model the large and computationally demanding systems of experimental data in a global approach.
- Local and Global Optimization algorithms: Another aspect of our work extends to the development of novel optimization algorithms for highly nonlinear multivariate systems. Our goal is to improve the stability and robustness of numerical optimization in for high-dimensional nonlinear problems. So far, we have successfully applied the use of genetic algorithms and a multidimensional spectrum analysis to obtain a multivariate optimization of finite element solutions to globally modeled sedimentation experiments. Mathematical modeling of hydrodynamic processes allows us to compare the efficiency, speed and robustness of both deterministic, gradient based approaches as well as stochastic approaches. Our approach is three-fold: First, we linearize a nonlinear problem in all dimensions, producing a sparse solution. This solution is further regularized by parsimonious regularization. Secondly, we use genetic algorithms to fit arbitrarily dimensioned nonlinear problems such as reacting systems. Thirdly, we employ Monte Carlo analysis to determine confidence limits for all parameters.
- High-Performance Grid Computing: An integral component of our research involves the application of high-performance computing (HPC) to accelerate computational intensive high-resolution analyses methods. We have built multiple supercomputer clusters at UTHSCSA and integrated them in the Texas Internet Grid for Research and Education (TIGRE) via grid middleware based on the Globus software stack. These resources have been developed to support our demanding computationally intensive modeling applications. Other HPC resources have been added in Munich, Germany (Leibnitz Rechenzentrum) and in Melbourne, Australia (VPAC and Bio21), as well as through NSF XSEDE. Analysis jobs can be submitted by registered users through a XSEDE Science Gateway, which provides computational resources through a Teragrid community account. This infrastructure is accessed with the UltraScan LIMS portal, a convenient web-based system for managing research data from analytical ultracentrifugation experiments.
- Rigid Body Bead Modeling: In collaboration with Dr. Mattia Rocco we are developing the SOlution MOdeler (SOMO) software which uses atomic level resolution models to derive bead replacement models to calculate hydrodynamic properties as well as modeling of small angle X-ray scattering profiles. Small groups of atoms (for example, in proteins, the backbone may be represented with one bead for each amino acid, while the sidechain is replaced with one or more beads, depending on size. The beads have the same average mass, center of gravity and approximate radius as the the atoms they replace. By preventing overlaps of these spherical beads by applying certain rules, an assembly of non-overlapping spherical beads can be used to predict hydrodynamic properties such as radius of gyration, sedimentation coefficient, translational diffusion coefficient and intrinsic viscosity on a macromolecule, which can then be compared to the observed hydrodynamic information from an AUC experiment. The SOMO software is integrated in UltraScan releases greater than version 9.9.
- Software Development: Our development team includes extensive expertise in software development. We employ modern open source tools to create a high-performance data analysis package for the management of research data from biophysical experiments. The software can be used to simulate and design experiments, analyze equilibrium and velocity sedimentation experiments, and includes a Laboratory Information Management System (LIMS) with a MySQL database back-end. Our expertise includes programming in C, C++, MPI, PERL, SQL, FORTRAN, PHP, HTML, Java, Python and other languages. Developments include portals and web interfaces to supercomputing applications, queuing systems, remote execution calls, XML procedures, and parallel file system storage technology. We follow standard programming procedures and employ open source tools such as Linx, gcc, and QT, QWT and QWT3D for our GUI frameworks.
- Experimental Design and Analysis: Our laboratory engages in a wide range of research projects involving proteins, protein assemblies, DNA, DNA binding protein complexes, nano particles, nano- and carbon tubes, colloids, interfaces, membrane proteins and other macromolecular systems. We design and conduct experiments to test scientific hypotheses and to research macromolecular structure and function, to characterize shape, and size distributions. To this end we engage in frequent collaborations with investigators at our institution and institutions across the US, Canada, Asia and Europe. We use the software and methodology developed in our laboratory to evaluate a large range of different systems, which also serves my goal for evaluating the validity and generality of the UltraScan software. Our expertise in this field is unique and results in important collaborative discoveries in biomedical sciences.