Search

The chemical structure information in patents remains difficult to access, partly because it is frequently expressed in the form of Markush structures, which can encompass enormous numbers of individual compounds. Early search systems were based on chemical ‘fragment codes’ that have still not been entirely superseded by the ‘topological’ systems developed during the 1980s. There are a number of databases of specific patented structures, which can be searched using standard substructure search software, and the more recently developed ones use automated data mining techniques to extract chemical nomenclature from patent text and translate it into searchable representations. Although some work has been done on automatic reconstruction of searchable Markush structures from patent text, this has proved to be considerably more refractory. A number of alternative approaches to chemical patent searching are being explored, some involving similarity and nearest-neighbor searching concepts, and some based on both existing curated databases and direct utilization of full-text patents. In-house systems, which facilitate integration with other cheminformatics systems, are also under development. These new systems may allow improvements in retrieval performance, especially with regard to search precision. © 2011 John Wiley & Sons, Ltd. WIREs Comput Mol Sci 2011 1 727-741 DOI: 10.1002/wcms.41

Most large pharmaceutical and biotechnology companies now use Oracle RDBMS chemistry data cartridges to manage databases of individual molecules for chemical structure searching. These systems are often linked to processes for new drug discovery and provide a common interface to a diverse range of specific structure databases. Recently some of these cartridges have been extended to handle Markush representations of un-enumerated combinatorial libraries alongside discrete molecules. An obvious extension would be to enable them to handle the Markush structures from chemical patents, though these have features and complexities not required for the representation of combinatorial libraries. The existing publicly available systems for handling patent Markush structures have changed little in the past 15 years and cannot easily be integrated with in-house systems; in-house access to chemical structures from patents is thus restricted at present to databases of specific molecules. A number of technical issues need to be tackled to enable the existing Markush-capable Oracle cartridges to handle data from patents, and several options are available for obtaining appropriate Markush structure databases for use with them. A demonstration system has been developed, using data from Thomson Reuters’ World Patents Index Markush File, and Digital Chemistry’s Oracle cartridge Torus. In-house access to patent Markush data could provide improved informatics support to the drug discovery process, both to enable patentability criteria to be added to computer-assisted drug design, and to expand the techniques available for data-mining in the patent literature.

HLA typing was performed on 384 individuals of an isolated population of 1,500 people with a.familial aggregate of lymphoma and immunodeficiency cases. Eighty-five % of the total population were descendants of the founding couple. First cousin marriages were common. There was a three-fold or higher increase of the following haplotypes as compared to the frequencies in Sheffield: HLA-A28, Bw35, HLA-A28, B18, HLA-A10, B18, HLA-A2, B18, HLA-A11, Bw40 and HLA-A11, B7. The frequency of HLA-A1, B8 was low (5.4%). The most common genotype was HLA-A2, B12/A2, B12 followed by HLA-A2, B12/A28, Bw35. We found 30 HLA homozygous individuals, of these 15 were HLA-A2, B12/A2, B12. There were two possible HLA cross-overs which may be confirmed and three postulated cross-overs which can never be confirmed as one or both parents of the individuals in question are deceased. Some of the haplotypes could be traced back to the first, second and third generations, i.e. to the first half of the nineteenth century. No single haplotype or antigen was shared by the patients.

We found HLA-DR5 to be present among 53% of 40 patients with goitrous autoimmune (Hashimoto's ) thyroiditis compared to 26% of 80 controls. No deviations from those expected in the incidences of HLA-A, B, C antigens were seen. In contrast HLA-DR3 was increased among 50 patients with atrophic thyroiditis (65%) compared to controls (24%). These findings stress the immunogenetic heterogeneity between the goitrous and atrophic varieties of thyroiditis.

One-hundred-and-forty-seven patients with autoimmune thyroiditis were studied with respect to HLA antigens as they related to various clinical features. HLA--B8 was found to be significantly increased among 59 patients with atrophic thyroiditis (57% vs. 26% for controls) but was identical to controls in 88 patients with goitrous thyroiditis (26%). No relation was found in either group between B8 and thyroid autoantibody titer or, in the case of goitrous thyroiditis, the rate of progression of the disease. Thus a link seems to be established between Graves' disease and atrophic thyroiditis in that both are significantly associated with HLA-B8. This study stresses the need to take clinical features into consideration when examining for HLA/disease associations.

This paper reports a comparison of calculated molecular properties and of 2D fragment bit-strings when used for the selection of structurally diverse subsets of a file of 44295 compounds. MaxMin dissimilarity-based selection and k-means cluster-based selection are used to select subsets containing between 1% and 20% of the file. Investigation of the numbers of bioactive molecules in the selected subsets suggest: that the MaxMin subsets are noticeably superior to the k-means subsets; that the property-based descriptors are marginally superior to the fragment-based descriptors; and that both approaches are noticeably superior to random selection.

This paper reviews the use of similarity searching in chemical databases. It begins by introducing the concept of similarity searching, differentiating it from the more common substructure searching, and then discusses the current generation of fragment-based measures that are used for searching chemical structure databases. The next sections focus upon two of the principal characteristics of a similarity measure: the coefficient that is used to quantify the degree of structural resemblance between pairs of molecules and the structural representations that are used to characterize molecules that are being compared in a similarity calculation. New types of similarity measure are then compared with current approaches, and examples are given of several applications that are related to similarity searching.

Though combinatorial libraries can be represented by enumerating the members of the library and using any of the commonly used computer representations for specific chemical structures, more compact forms are required for the processing of large ‘virtual’ libraries. Such representations can be divided into those which involve some form of ‘Markush’ or generic structure and those which use an approach based on generic reaction descriptions and sets of specific precursor molecules. This review discusses the nature of Markush structure representations, and the ways in which they are used in chemical patent information systems, and describes a number of different Markush-type representations developed especially for combinatorial libraries. Systems which utilise reaction-based library descriptions are also described, and the advantages and disadvantages of the two approaches, and the possibility of automatic interconversion between them, are discussed, along with the potential for their use in rapid analysis of library diversity.

Many different computer representations of chemical structures are used in drug design. Most treat the molecule as a topological graph, though the analogy is not a perfect one, and has problems with features such as tautomerism, aromaticity and stereochemistry. Commonly-used representations include 2D-structure diagrams, systematic nomenclature, line notations (e.g. SMILES), connection tables and the recently-developed International Chemical Identifier (InChI). Several approaches are used to indicate stereochemical configuration. 3-D structural representations are also used in identifying molecules with appropriate conformations for biological activity. The presence or absence of substructure fragments in a molecule is used to build chemical “fingerprints”, which are useful both in structure search systems and for measuring the similarity between molecules. It is frequently important to establish a unique representation for a molecule, and a number of canonicalisation algorithms and structure normalisation procedures (especially to deal with tautomerism and protonation) are used to achieve this. In some cases, these need to consider the predominant form under physiological conditions. “Business rules” for normalisation are especially important in chemical registration systems, which also need to deal with different salts and isotopically-labelled compounds, as well as unknown and partially-known structures. In recent years, many techniques have been developed for the analysis of structural databases; these include clustering, R-group decomposition, “reduced feature” representations and matched molecular pair analysis.

The history of the development of computer systems for storage and retrieval of Markush structures is reviewed briefly. The systems currently being developed by Chemical Abstracts Service, Derwent Publications/Questel/INPI, and International Documentation Company for Chemistry (IDC) are introduced, and the similarities and differences between the approaches they use for input representation and search are examined, especially in relation to the handling of generic nomenclature, the prospects for future development of such systems is discussed in light of recent and continuing research work, as is the potential for exchange of Markush structure databases between the different search systems.

The Molecular Information File (MIF) is a data-exchange format for chemical structure utilising the STAR File format described in Chapter 2.1. The current implementation addresses the core requirements for representing a two-dimensional molecular structure, with limited facilities for describing stereochemistry and the molecular geometry at stereogenic centres. The approach is complementary to that of the crystallographic information file (CIF) but the two formats use different subsets of the full STAR syntax and are therefore not fully compatible. A brief historical review is given of the evolution of MIF from the concepts of the Standard Molecular Data format of the late 1980s. After an account of the file structure and syntax, examples are given of representing atoms, bonds and molecules; of the ability of the format to accommodate a variety of bond-type assignments (including the simultaneous description of bonding by different schemes); the mechanism for describing and managing structural templates (e.g. ligands, functional groups, amino-acid units etc.); and stereochemical descriptors. There is also an account of the use of MIF as a database-query language. Keywords: MIF specification; Molecular Information File specification; chemical structure; molecular structure; save frames; looped lists; data blocks; global blocks; bonding conventions; stereochemistry; stereogenic geometry

This is version 1.2.3 of the Molecular Information File dictionary (MIF). MIF complements the Crystallographic Information File with data items describing two-dimensional chemical structure diagrams and bond types. The MIF format is described in Chapter 2.4.

One of the problems encountered in handling computer representations of combinatorial libraries (especially “virtual libraries”) is the extremely large number of compounds which may be covered by even quite simple libraries. This paper discusses work on the generation of structure fingerprints for the compounds in a library, which avoids the need to enumerate the compounds themselves by utilizing techniques originally developed for handling Markush structures in chemical patents.

Clustering is a data analysis technique that, when applied to a set of heterogeneous items, identifies homogeneous subgroups as defined by a given model or measure of similarity. Of the many uses of clustering, a prime motivation for the increasing interest in clustering methods is their use in the selection and design of combinatorial libraries of chemical structures pertinent to pharmaceutical discovery. One feature of clustering is that the process is unsupervised, that is, there is no predefined grouping that the clustering seeks to reproduce. In contrast to supervised learning, where the task is to establish relationships between given inputs and outputs to enable prediction of the output from new inputs, in unsupervised learning only the inputs are available and the task is to reveal aspects of the underlying distribution of the input data. Clustering is thus complemented by the related supervised process of classification, in which items are assigned labels applied to predefined groups: examples include recursive partitioning, naive Bayesian analysis, and K nearest-neighbor selection. Clustering is a technique for exploratory data analysis and is used increasingly in preliminary analyses of large data sets of medium and high dimensionality as a method of selection, diversity analysis, and data reduction. This chapter reviews the main clustering methods that are used for analyzing chemical

The commercially-available database systems for storing and searching Markush structures from chemical patents have undergone little change since their launch some twenty years ago. However, the past few years have seen the area become an active one again for research and development. This presentation offers an overview and commentary on recent and current activity, and discusses the prospects for improved access to structural information in the patent literature [1]. The existing curated Markush databases remain the gold standard, though several groups, both academic and commercial, continue to work on automatic analysis of full-text patents. This has involved not only the identification of specific-structure nomenclature and its conversion to structure-searchable records, but also the attempted reconstruction of searchable representations of complete Markush structures. The advantages and disadvantages of these approaches, and the prospects for their successful commercial exploitation, are discussed. New commercial software for searching Markush structure databases is being developed by several groups. These employ both conventional substructure search approaches (e.g. ChemAxon, Digital Chemistry), and novel algorithms, in some cases based on various forms of similarity and approximate structure matching (e.g. DecrIPt, IBM). These approaches are summarised and compared, and the opportunity their in-house implementation provides for integration of chemical patent information with the drug-discovery process is discussed. Practitioners have long been aware of the inadequacies and complexities of the existing systems, and the extent to which a new generation of systems may satisfy their requirements is discussed. The possible role of systematic evaluation of retrieval performance (in particular, the TREC-CHEM project) is addressed.

During the past few years interest has grown in the development of standard formats for the machine-readable presentation of chemical structures, and a few proposals have been published. One of these is the Standard Molecular Data (SMD) format, developed by a group of European chemical companies. Under the auspices of the Chemical Structure Association, a series of technical working groups have examined the original version of SMD format and proposed a number of revisions and extensions. This poster paper describes the revised version of the format using annotated examples, and discusses the areas where further extension is required.

The privatisation and reorganisation of British Telecom (BT) presented its various divisions with the problem of transferring commercial awareness at the top management level to middle management level. A management development programme was designed which combined computer‐based business strategy modelling and “best practice” modelling in human relations. These areas were previously treated separately. The greatest problem for senior managers was modifying their values. For middle managers the problem was that their values were overturned and themselves threatened. The one‐week courses allowed fears and confusions to be aired and examined. By exposing managers to “best practice” procedures they were given frameworks to help them determine how to forge a relationship between a sensible economic strategy and a complementary culture. The programme has been very successful and is to be implemented in other BT businesses.

A familial aggregate of seven cases of Hodgkin's disease (HD) has been investigated by HLA typing. Over 600 people in the immediate population (i.e. about half) have been HLA typed and haplotypes have been obtained for 95% of them. It was expected that the cases would share a particular HLA haplotype or at least that they would have one or two HLA antigens of the same series in common. However, this was not the case so no simple idea of association of HLA with HD cases was upheld. When antigen frequencies were examined in the whole population, it was found that HLA B18 increased progressively in incidence from 0.08 to 0.4 in successive groups of individuals each one more closely related to the HD cases. Similarly the community with the highest incidence of HD also had the highest incidence of B18. Thus B18, which in the world figures carries the highest relative risk, emerged as important in this study. Of four proposed interpretations of the data, we are most interested in the idea that the important HLA association is at a population level rather than at the level of the individual patient. A hypothesis, based on the concept of a "healthy carrier" for the HD agent, explains how such an association might operate. It is possible that B18-linked complement deficiency could be the basis for such a carrier state.

Mergers, takeovers and changes in products or markets confront managers with major cultural change in their organisations. This article looks specifically at British Telecom, the different demands made on managers pre‐ and post‐privatisation and describes now management development programmes helped managers to change their style.

Production Engineers have been continually seeking to find new and improved ways of converting an unmachined component into its finished design. Much of their time is spent investigating the contribution that new cutting tool materials and coolants could give them in the production of their components. In the field of production grinding more and more attention is being focused on the use of Superabrasives and in particular on ‘Cubic Boron Nitride’ for the grinding of ferrous materials. The purpose of this paper is to discuss the use of ‘Cubic Boron Nitride’ (C.B.N.) in the production grinding field and how the Engineer should approach the application of this highly versatile abrasive.