No abstract

No abstract

The Protein Data Bank is a computer-based archival file for macromolecular structures. The Bank stores in a uniform format atomic co-ordinates and partial bond connectivities, as derived from crystallographic studies. Text included in each data entry gives pertinent information for the structure at hand (e.g. species from which the molecule has been obtained, resolution of diffraction data, literature citations and specifications of secondary structure). In addition to atomic co-ordinates and connectivities, the Protein Data Bank stores structure factors and phases, although these latter data are not placed in any uniform format. Input of data to the Bank and general maintenance functions are carried out at Brookhaven National Laboratory. All data stored in the Bank are available on magnetic tape for public distribution, from Brookhaven (to laboratories in the Americas), Tokyo (Japan), and Cambridge (Europe and worldwide). A master file is maintained at Brookhaven and duplicate copies are stored in Cambridge and Tokyo. In the future, it is hoped to expand the scope of the Protein Data Bank to make available co-ordinates for standard structural types (e.g. a-helix, RNA double-stranded helix) and representative computer programs of utility in the study and interpretation of macromolecular structures.The Protein Data Bank' [1,2] was established in 1971 as a computer-based archival file for macromolecular structures. The purpose of the Bank is to collect, standardize, and distribute atomic co-ordinates and other data from crystallographic studies. As the number of solved protein and nucleic-acid structures has grown to the point where some lo7 characters are necessary to represent the co-ordinate information currently held, the need for such a computer-readable file has become very clear, and demands for the Bank's services have increased accordingly. The Protein Data Bank is one of several data base activities in the field of crystallography, e.g. the Bibliographic

Predicting and characterizing the crystal structure of materials is a key problem in materials research and development. It is typically addressed with highly accurate quantum mechanical computations on a small set of candidate structures, or with empirical rules that have been extracted from a large amount of experimental information, but have limited predictive power. In this letter, we transfer the concept of heuristic rule extraction to a large library of ab-initio calculated information, and demonstrate that this can be developed into a tool for crystal structure prediction.Ab-initio methods, which predict materials properties from the fundamental equations of quantum mechanics, are becoming a ubiquitous tool for physicists, chemists, and materials scientists. These methods allow scientists to evaluate and pre-screen new materials "in silico", rather than through time-consuming experimentation, and in some cases, even make suggestions for new and better materials [1][2][3][4][5]. One inherent limitation of most ab-initio approaches is that they do not make explicit use of results of previous calculations when studying a new system. This can be contrasted with datacentered methods, which mine existing data libraries to help understand new situations. The contrast between data-centered and traditional ab-initio methods can be seen clearly in the different approaches used to predict the crystal structure of materials. This is a difficult but important problem that forms the basis for any rational materials design. In heuristic models, a large amount of experimental observations are used in order to extract rules which rationalize crystal structure with a few simple physical parameters such as atomic radii, electronegativities, etc.. The Miedema rules for predicting compound forming [6], or the Pettifor maps [7] which can be used to predict the structure of a new binary material by correlating the position of its elements in the periodic table to those of systems for which the stable crystal structure is known, are excellent examples of this. Abinitio methods differ from these data-centered methods in that they do not use historic and cumulative information about previously studied systems, but rather try to determine structure by optimizing from scratch the complex quantum mechanical description of the system, either directly (as in ab-initio Molecular Dynamics), or in coarse-grained form (as in lattice models [8][9][10]). Here, we merge the ideas of data-centered methods with the predictive power of ab-initio computation. We propose a new approach whereby ab-initio investigations on new systems are informed with knowledge obtained from results already collected on other systems. We refer to this approach as Data Mining of Quantum Calculations (DMQC), and demonstrate its efficiency in increasing the speed of predicting the crystal structure of new and unknown materials. Using a Principal Component Analysis on over 6000 ab-initio energy calculations, we show that the energies of different crystal structures in ...

Our understanding of the classical MHC class I molecules (MHC class Ia molecules) has long focused on their extreme polymorphism. These molecules present peptides to T cells and are central to discrimination between self and non-self. By contrast, the functions of the non-polymorphic MHC class I molecules (MHC class Ib molecules) have been elusive, but emerging evidence reveals that, in addition to antigen presentation, MHC class Ib molecules are involved in immunoregulation. As we discuss here, the subset of MHC class Ib molecules that presents peptides to T cells bridges innate and acquired immunity, and this provides insights into the origins of acquired immunity.