| Project Members | Title |
|
Paul Brenner Dave Salyers Presentation Paper |
Proteins: Conformational Sampling and Flexible
Dihedrals Conformational sampling from X-ray and NMR structures is performed using our Shadow Hybrid Monte Carlo protocol, with modifications to include non-continuous motions of essential torsion angles. Analyses of dynamics and thermodynamics are used to characterize flexible and rigid residues. Dihedrals with slow motion are used as reaction coordinates to perform sampling and thus accelerate the sampling rate. We present test results for this method as applied to flexible protein-ligand systems. Protomol 2.0 Protomol is a high-performance framework in C++ for rapid prototyping of novel algorithms for molecular dynamics and related applications. Flexibility is achieved primarily through the use of object-oriented design strategies. Performance is obtained by using templates. In the cyclic design, implementation, and maintaince life cycle of Protomol we are currently between the design and implementation stages of Protomol 2.0. Protomol 2.0 refines the OO structure structure to further enhance the ease of framework understanding and new module addition. Example features include: reduction in the complications involved with adding an integration module, a more "GUI" friendly structure, simple io objects for building a fully robust set of input and output data formats, etc... For a final project students may review the new structure and select one or more modules from Protomol 1.0 which they would like to fully understand and then port into the new framework. Example modules key to an understanding of MD include: Particle Mesh Ewald, Shadow Hybrid Monte Carlo, and Equilibrium Molly. |
Scott Christley Xiaorong Xiang Presentation Paper |
Agent-based Simulation of Morphogenesis
We will design and implement an agent-based model and simulation of three-dimensional morphogenesis for cell sorting and chicken limb growth. Agent-based simulation is an alternative model to the Cellular Potts Model (CPM) previously used for morphogenesis; in contrast to CPM, agent-based models are not grid-based but can operate in any spatial representation and explicitly represent the cells in the model. An agent-based model consists of agents interacting with other agents and the environment within some spatial structure. The agents are the biological cells; the energy formulas used in CPM translate to the agent interactions, and morphogenesis is the emergence of global structure from the local interactions of many agents. The environment represents global interactions that are external to agents for processes like chemotaxis and haptotaxis. Agent-based simulations are generally considered slower than equation-based models because of the overhead associated with the simulation, but agent-based models have a direct conceptual link between entities in the model and entities in the real world phenomena that gives them a natural structure. By encapsulating entities, it may be possible to simplify agents, both in representation and functionality, yet still obtain the correct simulation results; or conversely, agents can be made more detailed without affecting the rest of the model. For the morphogenesis model, we will implement a simplified representation of a cell which uses less memory and more efficient algorithms for cell interactions. Besides performance, we will evaluate software engineering factors like modularity, maintainability, and expandability to see if agent-based models offer any advantages or disadvantages over CPM. |
|
Santanu Chatterjee Presentation Paper |
GRID-enabled Virtual Screening We will implement and test a virtual screening tool on the computational grid. We use a novel scoring function that approximates the binding free energy of protein-ligand complexes. The scoring function is obtained by computing the potential energy, the solvation free energy and entropy of the ligand, protein and protein-ligand complex. We validate our tool on the virtual screening of thousands of complexes with known experimental binding affinities using the computational grid. |
|
Trevor Cickovski Chengbang Huang Dave Cieslak Kedar Aras Presentation Paper |
BioLogo BioLogo is a domain-specific language developed for morphogenesis simulation engines. The BioLogo framework includes both a compiler which performs error-checking and converts BioLogo source into a temporary intermediate file, and a code generator which generates engine source, including any optimizations. BioLogo is currently being interfaced to a three-dimensional framework for morphogenesis simulation, CompuCell3D. We will attempt to add the generation of both CompuCell3D source code and Lua, and the implementation of more optimizations, to BioLogo. CompuCell 3D A combination Potts and PDE models can be used to model morphogenesis in multicellular organisms such as Dictyostelium Discoideum and Myxobacteria. The Dictyostelium slug, after forming a multicellular body from an initially unicellular state, follows a moving activator wave during early slug migration. The myxobacteria model exhibits a similar pattern, but enforces a polarity on cells that simulates bending. This will be accomplished through a polarity hamiltonian, adding an energy penalty making bending state favorable. Implementing irregular chicklimb shape in CompuCell3D CompuCell3D is a software for chicklimb growth simulation. This project is to implement the irregular shape of a chicklimb in CompuCell3D. A regular grid is required for Cellular Potts Model (CPM). First, embed the shape of a realistic shape into a regular grid; secondly, find a memory-efficient way to store the limb such that whenever a pixel of the regular grid is picked, it's fast to decide whether it's inside or outside the limb, or on the boundary of limb. Reference: A Framework for Three-Dimensional Simulation of Morphogenesis, T. Cickovski, C. Huang, R. Chaturvedi, T. Glimm, H.G.E. Hentschel, M. Alber, J. A. Glazier, S. A. Newman, J. A. Izaguirre, submitted, IEEE/ACM Transactions on Computational Biology and Bioinformatics, 2004. |
|
Michael Crocker Presentation Paper |
MDSimAid MDSimAid is a program written in Python whose purpose is to generate nearly optimal parameters for fast PME and MG fast electrostatics algorithms on a given input molecule. This is currently a work in progress. MDSimAid is located at http://www.mdsimaid.cse.nd.edu and a preprint can be found at http://www.nd.edu/~izaguirr/papers/CrHI0x.pdf. There are two main projects involving MDSimAid: MDSimAid assumes that the input molecule is in a state ready for simulation (correct structure, correct input files, minimized, and equilibrated). Extend MDSimAid such that the only input file needed is a PDB from the Protein Data Bank. All necessary system preperation is handled by MDSimAid. Add multiple time stepping (MTS) capability to the search routines in MDSimAid. |
|
Simon Kanaan Lance Gallop Presentation Paper |
Optimizing MSSC Optimize the program MSSC which predicts protein interactions. This program currently outperforms any other protein interaction predictor program currently available. These optimizations could include but not limited to changing the current data structure, matrix, to a more space/time efficient data structure such as a heap, modifying a few functions or redesigning the program in order to improve on the run time. Finally the last optimization would be to paralyze the code using for example MPI. Creating a Web Portal Create a web portal where a user can extract protein-protein interactions and protein-domain information from one or more data bases and run simulation using one of our protein interaction predicting programs such as MSC, MSSC, MESC. This web portal can consist of a GUI but need not be. |
|
Matt Rissler Presentation Paper |
Modeling Myxobacteria Aggregation in Compucell3D Myxobacteria aggregate into fruiting bodies when starved for an extended period of time. Myxobactria is an oriented rodshaped bacteria with a deØnite head and tail. The mechanism for this aggregation is head to tail C-signaling, and jamming. Lattice gas models of this system have been created by Mark Alber's group, as well as others. Since Compucell3D already has the framework for the extended Potts Model already in place, hopefully a myxobacteria simulation will be relatively easy to produce. A few of the issues that need to be addressed to realize a simulation of myxobacteria are: 1. Myxobacteria are rod-shaped, so some type of constraint to create stable rod-shaped groups of pixels is necessary. 2. C-signaling is a contact interaction, and can so be modeled as a near-neighbor interaction in the extended Potts model. However, it only occurs in a head to tail interaction. So a polarity of the cell must be accomplished. 3. Myxobacteria are not rigid bodies. Passive bending is important in aggregation, so must be possible in our simulations. |
|
John Tan Presentation Paper |
QFC Algorithm An algorithm called a quasi-periodic feature classifier (QFC) has been developed to have the ability to identify transmembrane proteins from genomic databases. Specifically, this QFC algorithm was trained to recognize G protein-coupled receptors (GPCRs). The algorithm and a brief background on GPCRs will be reviewed. Publications of subsequent applications of the QFC algorithm will be discussed. Background information and a study of potential GPCRs in the genome of the malaria-causing parasite, Plasmodium falciparum, are presented in this paper. 186 potential GPCRs were identified in the Plasmodium genome. Finally, the ability of the QFC algorithm to correctly distinguish between GPCRs and non-GPCRs is compared to Pfam profile HMMs (Hidden Markov Models). |