For the pharmacutical industry, developing a new drug is extremely costly and typically has a lead time of ten years. Parallelisation of the drug design software, MAXHOM, has speeded up the preclinical drug discovery and drug optimisation phases.
The MAXHOM consortium have broken new ground at the interface between genome research, pharmaceutical biology and computer science. genome research provides the basic information about medically important protein molecules; pharmaceutical biology attempts to produce new and more effective drugs targeted to these protein molecules; and computer science provides the tools.
The challenge lies in the enormous cost of drug development in the pharmeceutical industry. Currently the development of new drugs takes about 10 years, costing about 150 MECU. Competition is fierce, so the earlier a company can take a drug to market, the better its chances of success. Any delay in the drug development process leads to a drop in competitiveness and jeopardises the large initial R&D investment.
The problem is that genome research is one dimensional, a linear sequence of information symbols, while drug design uses 3-D shape information. The MAXHOM technology helps to close this gap by searching protein databases for information that relates 1-D sequences to 3-D shape.
MAXHOM makes considerable time savings in the first stage of drug design: the design process for leading compounds. Researchers are able to build 3-D models of the shapes of the protein molecules, the main targets of drugs. Such models are based on complex evolutionary similarities and it is these similarities MAXHOM is designed to detect in biological sequence databases. An important design goal was portability to a wide range of parallel machines, ranging from heterogeneous, wide area networks of workstations, up to massively parallel systems with distributed memory. As the parallel parts of MAXHOM are written in Fortran 77 using a standard message passing environment (PVM3, MPI will be supported in the future), the code can be easily installed on a great variety of parallel machines.
To circumvent a potential I/O bottleneck when the sequence database stored on disk is accessed from each process, a client-server model was developed. The achieved performance of a few minutes for a complete protein database scan means that one of the most sensitive database search algorithms, using a profile description of a protein family, can be performed in a nearly interactive fashion.
Sequence comparison has become an essential and standard tool in the analysis of genomic data. Genome projects will decipher much of the genetic information in many organisms, including humans. As a result, the computational cost of database searches will increase dramatically. In addition, the detection of sufficient similarity between a newly determined sequence to a protein of known function or even 3-D structure allows the transfer of most of the knowledge from one sequence to the other. This can result in enormous savings in genetic and biochemical laboratory efforts.
e-mail (email removed)
Research Area High Performance Computing and Networking
Project EUROPORT 2-Maxhom
Keywords drug design;parallelisation;
|E Merck DE|
|Parsytec Computer DE|
The URL for this page is /esprit/src/results/res_area/hpcn/hpcn15.htm
This page was last updated on 22 November 1996, and is maintained by (email removed)