Abstract
(Englisch)
|
Computational Biochemistry
We continue to operate our computational biochemistry web server (http://cbrg.ethz.ch) which has been consistently used by a larger number of bioscience labs throughout the world. Darwin, our language tailored for researchers in the biosciences, has also continued to expand with the creation of a more complete user manual, better on-line documentation, cleaner routines, and a more robust set of tools.
M. Hallett (Zurich) has produced a number of papers, reports, and software packages addressing problems from molecular evolution. Here we are concerned with evolutionary events beyond the standard insertions, deletion, and mutations. These events include gene duplications, gene reversals, recombination and horizontal gene transfers. A clearer understanding of the evolutionary events leading to extant sequences allows us to create reduced, purified datasets and gives us better data for use in our phylogenetic tree, multiple sequence alignment, and secondary structure prediction algorithms. Parts of this software are available via our server (http://cbrg.ethz.ch). Hallett has also worked on problems related to cell identification via mass spectrometry.
M. Hallett (Zurich) and A. Kahn (Zurich) have started a project entitled BioOpera with members of the Information and Communication Group here at the ETH. BioOpera is a process support system for bioinformatics. It allows users to design and specify a complicated set of interacting tasks that are required to complete a specific computation. It then schedules and loads balances this process over a cluster of machines in a persistent
fashion. During computation, BioOpera monitors the progress and records appropriate statistics. When jobs finish, the results are stored in a systematic way allowing for easy retrieval. When jobs fail, the system re-starts these tasks automatically. BioOpera complements our efforts with the programming language Darwin, since many algorithms for bioinformatic problems require vast amounts of computing time and space.
Computer Algebra
ALCOM-IT related work at ETH this year has focused on algorithms for solving linear systems over the ring of integers and ring of polynomials with coefficients from a field. A very fast algorithm for computing particular solutions to systems of linear diophantine equations has been developed. The algorithm is about an order of magnitude faster than than previous methods. The algorithm is eminently practical, one of the features being that ring extensions as required by previous approaches are avoided. Detailed pseudo-codes have been provided which should make implementation in a Computer Algebra system straightforward. The algorithm was presented by Thom Mulders at ISSAC'99 [1].
A related problem is to prove that a given system has no solution (when this is the case). More recently, we have extended the diophantine solver to produce such a certificate of inconsistency.
Reports
[1] Thom Mulders and Arne Storjohann. Diophantine Linear System Solving. In Proceedings of 'International Symposium on Symbolic and Algebraic Computation'. July, 1999. ACM Press. Vancouver, Canada.
|
