|
Jaeger (Biology) - The lab makes realistic conductancebased compartmental neuron models that match our electrophysiological data. Each neuron model is described by about 10,000 coupled ordinary differential equations. This complex system of equations needs to be solved numerically. GENISIS , a neuronal simulation software package is used for this purpose. We need to perform thousands of such simulations, as there is a need to search a very large parameter space to match the performance of the model with that of real
neurons. The cluster is essential to perform simulations of networks of such complex single neuron simulations. The benefits of this effort are tremendous: We can fully understand the interaction of multiple interacting nonlinear
properties of neurons, and we can simulate neuronal
computation.
Prinz (Biology) - Research in my lab uses experimental and computational approaches to study pattern generation and homeostasis in small neural circuits. The computational component of our work relies heavily on a tool I developed in the course of my postdoctoral research, namely the brute-force computational exploration of highdimensional parameter spaces associated with complex
biological systems. Because complex biological systems can have a large number of parameters, thorough exploration of parameter space can involve the simulation and analysis of several million parameter combinations. Without the computational power of the processor cluster that is partially owned by my lab this task would require prohibitive amounts of computation time.
Severson, Fu (Pharmacology) -
14-3-3 proteins are conserved regulatory molecules that are found in all eukaryotic cells, and which can bind to a diverse range of proteins including: kinases, phosphatases, and transmembrane receptors. These interactions
allow 14-3-3 to regulate many cell processes including cell-growth, signal transduction, and apoptosis. Using the crystal structure of 14-3-3 zeta bound to a
mode1 phosphopeptide, we have been using the highthroughput compute cluster to analyze potential binders to the 14-3-3 protein. This is done by taking 3D representations of individual compounds from a compound library (running them through DOCK 5.2, which calculates the most likely n modes of binding for a small molecule to a crystal structure. The high-throughput compute cluster is an ideal platform for this type of screening. The cluster sped up the computing
process for this screen from a timescale of months
|
