|
|
|
|
Computational Biology Cluster - Q2 1999 Status ReportThe Computational Biology Cluster (CBC) consists of 21 Pentium II workstations, 4 dual and 1 quad Pentium II servers connected by high speed 100Mbs Intel fast ethernet switches. The CBC has been made available to the biological computing community by remotely incorporating several of the dual workstations into the Theory CenterCluster where the scheduler is being developed. Thus far the cluster has been used to analyze multidimensional nuclear magnetic resonance spectra for determination of protein structure,bioinformatics analysis of protein DNA sequences together with protein structural data, Monte Carlo simulation of biomembranes, comparative analysis of molecular dynamics algorithms, and large scale parallel computation of protein and peptide structural hierarchies. Simulation of Biomembranes: (Feigenson Group) We are interested in how cholesterol interacts with other bio-molecules in the lipid bilayer of cell membranes. One way to study this interaction is to model the 2-dimensional molecular distributions using Monte Carlo simulation. In particular, we study the 2-dimensional cholesterol superlattices in model lipid membranes. We simulate the cholesterol lateral distribution with various interaction parameters, and calculate the chemical potential of cholesterol in the various environments. This study requires many repetitive, independent simulation runs. The CBC provides the ideal computation resource for this work. We discovered that cholesterol superlattices can not form with just pairwise additive interactions. Instead, a multibody interaction is required. We have characterized both the multibody form, as well as the magnitude of the interactions needed to produce cholesterol superlattices. Currently, the search for a possible physical origin of this multibody interaction is underway. Protein Structural Hierarchies (Shalloway Group) Our protein structural hierarchy calculations demonstrate the feasibility of large-scale parallel computation on high performance Pentium-based computers running Windows NT and using the Parallel Virtual Machine (PVM) message passing library. These calculations made use of 40 processors and ran continuously for over 72 hours generating data sets of over 100 GB per run. These large data sets are stored on hard disks distributed throughout the cluster and are accessed by a distributed data manager that also uses PVM. |
|
Last modified on: 10/12/99 |
