Iñaki Martinez de Ilarduya Muñoz
Resumen de la Memoria Crystallographic methods exploiting modern computing.
SUBIX, SHREDDER, middleware and web interfaces, tools for crystallographic teaching Structural biology has experienced a rapid development during the last years, as can be seen in the statistics describing the growth in the number of entries deposited with the public repository PDB (http://www.rcsb.org/pdb/statistics/contentGrowthChart.do?content=total&seqid=100). The pdb holds information for protein and nucleic acid structures as well as for complexes involving both kinds of macromolecules. The structures deposited are the product of diverse experimental or modeling techniques but currently single crystal X-ray diffraction is not only responsible for the vast majority of the studies (89%), it also yields the highest level of detail and precision. This yearly accelerating pace in the solution of new crystal structures has been made possible by the development of new algorithms and techniques, boosted by the advances in modern computing resources available to the scientific community. This is precisely the scope of the work underlying the present thesis: developing methods and exploiting techniques aiming to maximize the performance of computing resources and bringing in modern computing tools. Paradoxically, even though the crystallographic community has been relying on the use of computers ever since their early incorporation as tools in universities and research centers over five decades ago, it has largely turned its back on modern supercomputing. Indeed, crystallographic software rooted in these primitive systems has been thrifty in the use of computing resources, producing highly optimized programs able to run fast on any available equipment. While other fields, notably all the ¿omics¿ such as proteomics, genomics, interactomics¿ have resorted to the use of distributed systems and highly parallelized algorithms, this angle has been lost to crystallography. Supercomputing is perceived as an unnecessary complication by the crystallographic community but computers in a network can be exploited in a concerted way and made to work as flexible supercomputers. The increase in computing power allows to propose and exploit new algorithms and methods based on massive calculations, as long as the job is divided into smaller tasks that can be separated and distributed, after the ¿divide and conquer¿ principle. Aggregation of the results leads to the conclusion in affordable time. Finally, methods developed on the basis of distributed systems can be upscaled to a supercomputer and made available through user friendly implementations providing easy to use solutions. In this frame, the scope of the present work has been set on the development of computing tools exploiting modern technologies through a variety of aspects:
¿ Development of a program conceived to identify models for Molecular Replacement in nucleic acid structures exploiting the advantages of current databases.
¿ Development of a methodology to generate and use protein fragments derived from distant homology models for Molecular Replacement.
¿ Modern computing tools for crystallography o Application of GRID computing to existing tools and for those designed to exploit it, adapting algorithms and methods to exploit the advantages of accessing multiple computers in parallel, to target crystallographic problems.
o Web interface for secure and efficient management of the programs developed in our research group, user accounting and download site.
¿ Development and implementation of a portable, practical tutorial concept designed to teach crystallographic computing methods as well as for the dissemination of our software among users.
The present thesis is comprised of four chapters describing each of the aforementioned objectives.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados