Data processing: structural design – modeling – simulation – and em – Modeling by mathematical expression
Reexamination Certificate
2000-01-14
2004-01-06
Moran, Marjorie (Department: 1631)
Data processing: structural design, modeling, simulation, and em
Modeling by mathematical expression
C703S012000, C702S027000
Reexamination Certificate
active
06675136
ABSTRACT:
The present invention relates to a computer based technique for the global investigation of property spaces. More particularly, but not exclusively, the present invention relates to a computer based technique for the global investigation of chemical space, with reference to drugs, using a reference set of molecules and descriptors that allows the systematic mapping of the global chemical space. The technique, may be used, for example, to generate a global map of the multidimensional chemical space that allows one to examine, in a consistent manner, the inner relationship between various molecules.
Drug discovery is a time and resource consuming exercise. Current computer based tools allow the description of chemical spaces as local models. Many chemical fields have been targeted by approaches used, or proposed, to discover new chemicals. For example, pharmaceuticals, agrochemicals, cosmetics and perfumes, photographic materials and others have benefited from the methodology developed to assist chemical synthesis. In all these fields, central to the goal of discovery is the novelty of chemical structures, and the novelty of chemical properties. With the advent of parallel synthesis and combinatorial chemistry, large numbers of chemical compounds are now within reach for synthesis and evaluation. Crucial for the practising chemist remains the goal of prioritising, out of thousands or millions of possibilities, which compound to make next.
For a pharmaceutical or agochemical compounds, there are two main types of relevant information in a molecule, i.e. chemical and biological. Medicinal chemistry handles chemical information by identifying classes of “active molecules”, then zooms in on biologically relevant information by performing various bioassays. However, in the initial stages of a research project, where little or no information is available concerning the biological target, chemical information is the only property that one can handle appropriately. Increasing the chemical information known about each compound becomes a goal of such early-phase projects, especially in the absence of active compounds.
“What's the best way to describe a molecule numerically and uniquely? What's the best way to categorise clusters of molecules? Is all this work producing results that are any better than plain old random selection?” These questions are quoted from an article by Elizabeth K. Wilson, “Computers customise combinatorial libraries”, published Apr. 27 1998 in Chemical & Engineering News (pp. 31-37). This article summarises the issues discussed at the recent American Chemical Society “Diversity Symposium”, organised by Robert S. Pearlman in Dallas, Tex. According to this article, the issues of molecular diversity, and of describing chemicals in an unique and relevant manner, have not been resolved. There is no general consensus as to which approach should be taken.
Molecular similarity is an ubiquitous concept that originates from the XIXth century. Attempts to rigorously define molecular similarity can be found in the book “Concepts and applications of molecular similarity”, edited by Mark A. Johnson and Gerald M. Maggiora, J. Wiley & Sons, ISBN 0-471-62175-7, 1990. The impact of molecular similarity in the field of drug design, and a survey of recent advances of using molecular similarity in the pharmaceutical industry have been the aim of the book “Molecular similarity in drug design”, edited by Philip M. Dean, Chapman & Hall, ISBN 0-7514-0221-4, 1995.
Molecular diversity has been the target of recent molecular similarity-based methods, in the effort to maximise the structural diversity of combinatorial and/or HTS libraries, so as to ensure the largest possible coverage of the chemical space. Molecular diversity analysis methods are surveyed in volume 7/8 of Perspectives in Drug Discovery and Design, ISSN 0928-2866, “Computational methods for the analysis of molecular diversity”, edited by Peter Willet (1997).
Presently, tools to describe chemical space are used to generate local models. For example. Sergio Clementi and co-workers have described the principal properties space for a set of 40 heteroaromatic compounds in Quant. Struct.—Act. Relat. Vol. 15, pp. 108-120 (1996). In their work, Clementi et al. calculated various properties for 45 compounds running the GRID-programme, which is based on three-dimensional descriptors. Out of the resulting calculations, they have derived a set of principal properties, and have classified these compounds into ten clusters. However, they classified guanine, a biologically important heteroaromatic ring, as an “outlier” that falls outside the property space of the aforementioned GRID descriptors. One of the inherent limitations of local models is that the validity of the analysis is only as good as the dataset composition, and unique features are reflected into “outliers” which often tend to skew the statistical results and are, therefore, excluded from the analysis.
Furthermore, local models tend to be outdated, as new data are generated. This is illustrated by work performed by Svante Wold and co-workers where an initial three-dimensional property space for the 20 natural amino acids, J. Med. Chem., vol. 30, pp. 1126-1135 (1987), was extended to a five-dimensional set for 55 amino acids, Quant. Struct.—Act. Relat., vol. 8, pp. 204-209 (1989). Recently, this was further extended to a set of 87 amino acids, still using a five-dimensional property space, and published in J. Med. Chem., vol. 41, pp. 2481-2491 (1998). The 5 principal properties derived for amino acids are similar to the Hammett and Taft parameters, widely used in physical organic chemistry textbooks to correlate physico-chemical properties with molecular structures. These properties, termed “Z-scales”, have been tentatively interpreted as measures of lipophilicity (z1), size/polarizability (z2), polarity (z3), while the fourth and fifth scales (z4 and z5) were more difficult to interpret. This work has extended the principal property space represented by the twenty natural amino acids with an additional set of 67 non-coded amino acids, some of them explicitly synthesised to cover unique properties. However, these Z-scales remain valid only for amino acids, and further synthesis of novel structures would lead to revaluation of the principal properties, and of the “Z-scores” for individual amino acids.
Current computer based technology allows the end-user to generate, in silico, extremely large numbers of compounds. For example, Tripos Inc. and Silicon Graphics have announced that they in a joint project have created a virtual library consisting of 100 billion molecules, using a “SpaceCrunch” technology.
Tripos' software ChemSpace™ yields “all possible molecular products resulting from given reactions, allowing chemists to start travelling with confidence over large expanses of the chemical universe”. This software promises a structural description of the chemical universe/space, based on single compounds and within certain limits. Chemspace™ is a searchable database consisting of billions of compounds synthesizable from known reactions and available reagents. This method includes tools to navigate in the database. However, the database represents only a subset of the chemical space, limited by the type of chemical reactions and reactants provided in “SpaceCrunch”. This stepwise manner to map chemical space has been, so far, the only alternative to true chemical space navigation.
From all the above, one can observe that there is a considerable need to navigate in the chemical space.
The present invention addresses the disadvantages discussed above and allows one to generate a global model that includes, and can specifically analyse, (e.g. heteroaromatic compounds (vide infra)) without the risk of extrapolation or outlying behaviour, given that the raw data are correct.
It is an object of the present invention to provide a computer based method to investigate any property space, e.g. a chemical and/or biological space, based on a set of objects (st
Gottfries Johan
Oprea Tudor
AstraZeneca AB
Moran Marjorie
White & Case LLP
LandOfFree
Global method for mapping property spaces does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Global method for mapping property spaces, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Global method for mapping property spaces will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3250687