This invention relates to computational chemistry methods, systems and computer program products, and more particularly to methods, systems and computer program products for identifying pharmacophores in molecules.
The discovery of new chemically active and/or bioactive molecules has become increasingly important, for example in drug discovery. Unfortunately, however, the binding of a molecule into a protein generally is inherently a three-dimensional event. Moreover, only about one third of the current drug targets presently have a crystal structure of the target protein to guide drug design efforts. Accordingly, it is desirable to provide computational methods that can use biological assayed potency data and/or other empirical data to identify one or more pharmacophores, i.e., an arrangement of key binding features in space. It may also be desirable to know a binding conformation and the key features that allow a molecule to bind to a protein. As used herein, a “conformation” means any of the spatial arrangements of a molecule that can be obtained by rotation of the atoms about a single bond, and a “feature” means two or more molecular elements and the related distances therebetween. Once a pharmacophore is identified computationally, it can be used to guide selection of compounds to screen, virtual screening, and/or to guide lead optimization efforts, the modification of existing molecules and/or the design of new molecules.
In view of the above, considerable effort has been expended to infer pharmacophore features using only biological potency of molecules along with their crystal structures. Pharmacophore identification methods, systems and computer program products are described, for example, in Martin, “Pharmacophore Mapping”, Chapter 6 in Designing Bioactive Molecules: Three Dimensional Techniques and Applications, American Chemical Society, Washington, D.C., 1998; Güner, Ed., Pharmacophore, Perception, Development, and Use in Drug Design, International University Line. La Jolla, Calif., 1999; Kurogi et al., Current Medicinal Chemistry, 9, pp. 1035-1055, 2001; Patel et al., J Comput Aided Mol Des., 16, pp. 653-681, 2002; and Güiner et al., Current Medicinal Chemistry, 11, pp. 2991-3005, 2004. Pharmacophore identification is also described in U.S. Pat. No. 6,421,612 to Agrafiotis et al., entitled System, Method and Computer Program Product for Identifying Chemical Compounds Having Desired Properties, and published U.S. patent application Ser. No. 2004/0236555 A1 to Pierce et al., entitled Target Ligand Generation. See also, U.S. Pat. No. 6,434,542 to Farmen et al., entitled Statistical Deconvoluting of Mixtures, and U.S. Published patent application Ser. No. 2003/0061186 to Farmen et al., entitled Statistical Deconvoluting of Mixtures.
Pharmacophores in a plurality of molecules may be identified, according to various embodiments of the present invention, by generating a set of conformations for a respective molecule. A respective conformation comprises a series of features that are present or absent in the conformation, wherein a respective feature comprises at least two molecular elements and at least one distance therebetween. The features for a set of conformations for a given molecule are repeatedly compared to a model of feature importance of remaining molecules, to identify an inferred conformation for the given molecule and update the model of feature importance, until the model of feature importance for the plurality of molecules converges. In some embodiments, the repeated comparing of the features for a set of conformations for the given molecule to a model of feature importance of the remaining molecules is performed using Gibbs sampling, expectation maximization, stochastic sampling and/or other techniques.
In some embodiments, the repeated comparing of the features for a set of conformations for the given molecule to a model of feature importance of the remaining molecules, to identify an inferred conformation for the given molecule, until the model of feature importance for the plurality of molecules converges may be performed by selecting an initial inferred conformation for a respective molecule from the set of conformations for the respective molecule, to generate an initial set of inferred conformations for the plurality of molecules. Then, for a given molecule, features that are present in the inferred conformations for the remaining molecules are identified to obtain an initial model of feature importance. The set of conformations for the given molecule is then compared to the initial model of feature importance to identify an updated inferred conformation for the given molecule. Then, the initial inferred conformation for the first molecule is replaced with the updated inferred conformation for the first molecule. The above operations are then repeated for the remaining molecules, to obtain an updated model of feature importance and a set of updated inferred conformations for the plurality of molecules. The above operations are repeatedly performed until the updated model of feature importance for the plurality of molecules converges, to thereby obtain a final model of feature importance.
In some embodiments of the present invention, the identifying of features that are present in the inferred conformations for the remaining molecules to obtain an initial model of feature importance is performed by identifying features that are present in the inferred conformations for the remaining molecules to obtain a model of feature prevalence. Then, the model of feature prevalence is compared with a model of historical prevalence for the features to obtain the model of feature importance. Finally, the conformations for the given molecule are compared with the model of feature importance to obtain a model of conformation likelihood for the respective conformations of a given molecule.
In still other embodiments of the present invention, the model of feature importance is repeatedly mapped onto the respective inferred conformations, to obtain one or more pharmacophore hypothesis sets of features for a respective conformation. Inferred conformations that are consistent with the respective hypothesis sets of features are identified. One or more of the pharmacophore hypothesis sets is then selected based upon a measure of conformations that are consistent with a respective pharmacophore hypothesis sets. In some embodiments, repeated mapping of the model of feature importance onto the respective inferred conformations is performed by identifying one or more completely connected graphs that are consistent with the respective hypothesis sets of features that were inferred. The one or more completely connected graphs may be identified by executing a clique finding algorithm, such as a Bron-Kerbosch algorithm. Accordingly, some embodiments of the present invention can identify two or more pharmacophores that are consistent with the inferred conformations that have converged.
Moreover, the identification of multiple pharmacophores in a given data set according to embodiments of the present invention may be used independent of the particular method that is used to identify pharmacophores. In particular, some embodiments of the present invention can generate a set of conformations for a respective molecule, wherein a respective conformation includes a series of features that are present or absent in the conformation, and a respective feature includes two or more molecular elements and one or more distances therebetween. Two or more pharmacophores are then identified that are consistent with the set of conformations. If there are two pharmacophores, then the identified conformations of one set of molecules may give rise to one pharmacophore and the identified conformations of another set of molecules may give rise to the second pharmacophore. The two or more pharmacophores may be identified by identifying two or more completely connected graphs that are consistent with the set of conformations for each set of molecules. The connected graphs may be identified by executing a clique finding algorithm, such as Bron-Kerbosch.
Still other embodiments of the present invention can compare the set of conformations for the given molecule to the initial model of feature importance to identify an inferred conformation for the given molecule by comparing, using regression techniques, the set of conformations for the given molecule to the initial model of feature importance, taking into account relative potency data for the plurality of molecules, to identify an updated inferred conformation for the given molecule. Comparing using regression can be performed using multiple linear regression, partial least squares, ridge regression, neural network, logistic regression and/or support vector machine techniques.
Moreover, this comparing using regression may be used to identify pharmacophores in a plurality of molecules from a set of conformations for a respective module, by comparing the set of conformations with the respective molecules to a model of feature importance, taking into account relative potency data for the plurality of molecules, to identify inferred conformations for the plurality of molecules. Accordingly, relative potency data may be taken into account in identifying pharmacophores, using regression.
It will be understood that embodiments of the invention have been described above primarily with respect to methods of identifying pharmacophores. However, analogous systems and computer program products also may be provided.
The present invention now will be described more fully hereinafter with reference to the accompanying drawings, in which illustrative embodiments of the invention are shown. However, this invention may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
It will be understood that when an element is referred to as being “coupled”, “connected” or “responsive” to another element, it can be directly coupled, connected or responsive to the other element or intervening elements may also be present. In contrast, when an element is referred to as being “directly coupled”, “directly connected” or “directly responsive” to another element, there are no intervening elements present. Like numbers refer to like elements throughout. As used herein the term “and/or” includes any and all combinations of one or more of the associated listed items and may be abbreviated by “/”.
It will also be understood that, although the terms first, second, etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms “a”, “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises,” “comprising,” “includes” and/or “including” when used herein, specify the presence of stated features, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and/or groups thereof.
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art in light of the present disclosure, and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
The present invention is described in part below with reference to block diagrams and flowcharts of methods, systems and computer program products according to embodiments of the invention. It will be understood that a block of the block diagrams or flowcharts, and combinations of blocks in the block diagrams or flowcharts, may be implemented at least in part by computer program instructions. These computer program instructions may be provided to one or more enterprise, application, personal, pervasive and/or embedded computer systems, such that the instructions, which execute via the computer system(s) create means, modules, devices or methods for implementing the functions/acts specified in the block diagram block or blocks. A computer program according to embodiments of the invention comprises a computer usable storage medium having computer-readable program code embodied therein. Combinations of general purpose computer systems and/or special purpose hardware also may be used in other embodiments.
These computer program instructions may also be stored in memory of the computer system(s) that can direct the computer system(s) to function in a particular manner, such that the instructions stored in the memory produce an article of manufacture including computer-readable program code which implements the functions/acts specified in block or blocks. The computer program instructions may also be loaded into the computer system(s) to cause a series of operational steps to be performed by the computer system(s) to produce a computer implemented process such that the instructions which execute on the processor provide steps for implementing the functions/acts specified in the block or blocks. Accordingly, a given block or blocks of the block diagrams and/or flowcharts provides support for methods, computer program products and/or systems (structural and/or means-plus-function).
It should also be noted that in some alternate implementations, the functions/acts noted in the flowcharts may occur out of the order noted in the flowcharts. For example, two blocks shown in succession may in fact be executed substantially concurrently or the blocks may sometimes be executed in the reverse order, depending upon the functionality/acts involved. Finally, the functionality of one or more blocks may be separated and/or combined with that of other blocks.
As also shown in
Continuing with the description of
The data structure 134 of
The data structure 134 of
Referring to
Still referring to
Continuing with the description of
In particular, at Block 610, the model of feature importance Q (210) is repeatedly mapped onto the respective inferred conformations (and hence molecules) 250, to obtain one or more pharmacophore hypothesis sets of features for a respective conformation in the set. At Block 620, inferred conformations 250 that are consistent with the respective hypothesis sets of features are identified. Finally, at Block 630, one or more of the pharmacophore hypothesis sets is selected based upon a measure of conformations (and hence molecules) that are consistent with the respective pharmacophore hypothesis sets.
Still referring to
Additional discussion of various embodiments of the invention that were described above in connection with
In particular, small molecules typically have rotatable bonds and often have flexible rings so they can take any of a vast number of shapes. Each shape will put atoms into different relative positions in space. It is desirable to know the binding conformation for each molecule. If there are multiple binding modes or binding sites, it is also desirable to know which molecules may be put together as they are binding in the same way. It is also desirable to identify the key features that interact with the target protein or, if the assay system contains multiple targets, the molecules that bind to each target. Both the determination of the binding conformation and its key features can have extremely large search spaces. For example, a molecule with five rotatable bonds and three flexible six member rings might be represented with 3523=1944 distinct conformations. Much effort has been expended in selecting features to represent the biologically important characteristics of a molecule, as described, for example, in Leach et al., An Introduction to Chemoinformatics, Kluwer Academic Publishers, Dordrecht, 2003.
One technique of feature representation is to create a bit string where each bit gives the presence or absence, for example 1/0, of a feature in the conformation of the molecule. There are a number of features that medicinal chemists have considered important. A molecular feature can include two atoms or molecule fragments and the through-space distance between them. Typical features include the following: hydrogen bond donor, hydrogen bond acceptor, positively charged group, negatively charged group, lipophilic group, aromatic ring center, halogen, etc. The distance is often given in a range, e.g. three to five Ångstroms. Thus, a feature might be HBD 3-5 HBA. The features and the distance bins of 2-4, 4-8, 8-12, 12 or more, may give rise to 105 pharmacophore-pair/distance features. Using these features, each conformation of a molecule can be represented as a bit string of 105 bits. A richer list of features, more distance bins, or extensions to three, four or five features can vastly increase the number of bits in the string used to represent a conformation.
Some embodiments of the invention can use a modification of Gibbs sampling (described, for example, in Lawrence et al., Science, 262, pp. 208-214, 1993, and Rouchka, sapiens.wustl.edu/˜ecr/PAPERS/gibbs.pdf, 1997) whereby the most likely binding conformations and the key binding features are identified for each molecule. In some embodiments, these results are used with a clique detection method such as Bron-Kerbosch (as described, for example, in Bron, Comm ACM, 16, pp. 575-577, 1973), find a rigid set of features and with fixed distances between them.
Some embodiments of the present invention may provide four general analysis operations: feature definition, pharmacophore fingerprint alignment, hypothesis generation and refinement. Feature definition was described generally above at Block 310. Pharmacophore fingerprint alignment was described generally in Block 320, and more specifically in
1. Feature Definition
In some embodiments, six pharmacophore features are predefined: hydrogen bond donor, hydrogen bond acceptor, aromatic ring center, hydrophobic center, negative charge center, and positive charge center. Certain features can be defined as a vector, which means the angle between direction or normal vectors satisfies a predefined threshold.
In some embodiments, the well known Daylight SMART language is used to define pharmacophore features. Users can change the rules based on their understanding of the problem. For certain datasets, like ACE inhibitors, this may be desirable for pharmacophore identification since there are some uncommon definitions for pharmacophore features, like zinc binding site, in the ACE dataset.
The distance between each pair or pairs of features is computed. The distances are placed into bins. In some embodiments, the default bins are 2-4, 4-8, 8-12, 12 or larger Ångstroms. The user can change the definitions of the bins, add more bins, overlap bins, etc. During pharmacophore generation stage, the selected conformations for all molecules may be superimposed, and the actual distance ranges may be calculated if the superimposition satisfies a user provided Root Mean Square Distance (RMSD) cutoff. The features and the observed distances between them may be used to define the final pharmacophore.
A feature pair is defined as two pharmacophore features and the distance (binned) between them. If a conformation has the pair, a certain bit is set to one. If a conformation does not have the pair, that bit is set to zero. A bit string is used to represent each conformation. As shown in
2. Pharmacophore Fingerprints Alignment
A general framework that encompasses many embodiments of pharmacophore fingerprint alignment now will be provided. An abstract formulation of the pharmacophore identification problem is as follows. From a set of conformations of several compounds, find:
An appropriate objective function F(P,C) is formulated that provides a numerical evaluation of how good a particular choice of P and C is, and then the values of P and C are found that can maximize this function. Finding the optimal values of P and C may be difficult since the P-C space of potential solutions may be prohibitively large for systematic exploration. Embodiments of the invention can reduce the size of the solution-space by proceeding iteratively, alternating between P and C. For example, fix the current value of P and then select an appropriate C in light of the current P, then fix the C and select another P, and so on. A selection scheme is devised that can lead to optimal (or at least good) solutions for P and C if the iterative process is carried out long enough. Selection schemes according to some embodiments of the invention can fall into at least three broad categories:
Some embodiments of the invention may provide a composite of all the three schemes described above with a hybrid scheme forming the dominant component. Greedy maximization may be used to find good starting points and approximate solutions. In some cases where there is clear signal in the data, this scheme may be adequate by itself. Pure stochastic sampling may be applied in cases when the solutions found using a combination of the other two schemes appear unsatisfactory. It may also be used in more complex situations—such as when multiple binding modes might be involved.
In addition to selecting the scheme or combination of schemes, an implemented embodiment may also involve the selection of:
Reference implementations to demonstrate embodiments of the invention and to study their effectiveness may include two computer programs—one is a pure greedy maximization approach, and the other uses a hybrid strategy. Both programs can use the probability models presented in the above-cited Lawrence, et al. publication. The objective function F corresponds to the “likelihood function”. Lawrence, et al. devised a hybrid scheme for the problem of genomic sequence-alignment. Their problem involved probability models and some algorithmic issues that may be adapted to binding conformation problems according to embodiments of the invention (which has a different data structure). They also proposed a greedy maximization version of the algorithm which can be used in embodiments of the invention. Lawrence et al. referred to their algorithm as a predictive Gibbs sampler and used the term “Gibbs sampling” as notational shorthand for the algorithm.
3. Hypothesis Generation
Fingerprint alignment picks out one conformation for each compound and a series significant pharmacophore features, in the case feature pairs (i.e., two point pharmacophores). However, whether these conformations can be combined to have common 3 points, 4 points or even higher order pharmacophores also may be determined according to some embodiments of the present invention.
After the hypothesis generation (Block 610), a list of pharmacophore hypotheses is created. Every hypothesis can be evaluated in the refinement stage.
The following examples shall be regarded as merely illustrative and shall not be construed as limiting the invention. Several examples will be used to describe embodiments of the invention. First, a toy example of synthetically constructed bit strings will be given. Next, it will be shown how embodiments of the invention can find the binding conformation and some of the key features for a small set of D2 ligands. Embodiments of the invention then are used to examine a large data set of ACE inhibitors. Two pharmacophores are found that utilize 65 and 57 molecules from the set of 78 ACE inhibitors. Finally, selected D2 and D4 active compounds (described, for example, in Bostrom et al., J Chem. Inf Comput. Sci., 43, pp. 1020-1027, 2003) are used to mimic a situation where there are two distinct binding modes or sites. All conformations were computed using Omega from OpenEye Scientific Software (as described, for example, in Omega, eyesopen.com/products/applications/omega.html).
Example 0. To demonstrate that embodiments of the invention can successfully identify pharmacophoric features a toy data set of 0's and 1's was build. There are 20 molecules, each with 20 features and 20 conformations. The incidence of 1's was set low, ˜5%, to mimic the sparse nature of bit string representations of conformations. A conformation for each molecule was selected at random and bits were set to 1 for positions 1, 5, 10, and 18. So each “molecule” has one “binding” conformation with four features. Embodiments of the invention successfully found the binding conformation for each molecule.
Example 1. A 3D study of D2 and D4 ligands was undertaken by Böstrom et al. (as described in the above-cited Böstrom publication). They report a 3D pharmacophore. Twenty one (21) of the more active D2 compounds were processed according to embodiments of the invention and reproduced their results. See
Example 2. A total of 78 ACE inhibitors taken from DePriest et al., J. Am. Chem. Soc. 115, pp. 5312-5384, 1993, with a total of over 46,000 conformations, were processed according to embodiments of the invention. Sixty-five of the compounds satisfied the pharmacophore shown in
Example 3. A 3D study of D2 and D4 ligands, as described in the above-cited Bostrom et al. publication, reports 3D pharmacophores for D2 and D4 activity. Six of the compounds that were relatively selective for D2 compounds were submitted, together with six of the compounds that were relatively selective for D4. Both types of compounds were submitted together to mimic a situation that might be expected to happen in practice where the experimenter is unaware that there are two binding modes or locations. Embodiments of the invention reproduced the pharmacophores for both D2 and D4. See
Additional discussion will now be provided. One way to think about the pharmacophore identification problem is to consider the most potent and rigid molecule and determine if other molecules and their features can be mapped onto this pharmacophore prototypical molecule. It can be a very big intellectual and computational leap to consider all the molecules. Even pair-wise comparisons among N molecules generally increase at a rate proportional to N2. Moreover, each molecule will typically have many energetically feasible conformations. The full-scale problem, the comparison of each conformation of one compound to all the conformations of all the other compounds, may be essentially computationally hopeless. For example, 500 conformations each of 100 molecules may lead to extremely large numbers of possibilities. Some embodiments of the invention, however, can create a model of the active conformation and its key features, the Q vector and A matrix. The information about this model is captured in the Q vector that gives the probability that a feature is important and the A matrix giving the probabilities that a conformation is the binding conformation. Given the model of the active conformation and its key features, each conformation for each molecule under consideration can be compared to the model. What would have been a many to many comparison, now can become a many to one comparison, according to some embodiments of the invention.
It appears somewhat remarkable that some embodiments of the invention can start with essentially nothing (neither the active conformation for each molecule nor the key features are known) and that some embodiments of the invention can converge to a good answer. Statistical theory may imply that embodiments of the invention will converge, but appears silent on the rate at which embodiments of the invention will converge. Embodiments of the invention appear to work on the above examples and may be expected to work on data sets where the data is consistent with a solution. A potential justification on why embodiments of the invention should work is as follows. Start with a randomly chosen conformation for each active molecule. Search these conformations for features that are in common. Once a few features that are in common (and presumed to be important) are found, embodiments of the invention can judge better which conformations are likely to be binding. Given better conformations, embodiments of the invention are then better able to point to binding features. In practice, in the examples studied, the process appears to converge rapidly to a solution.
A data set can be trivial. Suppose that the data set contains a very active, rigid molecule. If there is a rigid compound then an active analog method may be used, as described, for example, in Marshall et al., “The Conformational Parameter In Drug Design”, in Computer-Assisted Drug Design, Olsen, E. C. Christoffersen, R. E. Eds., American Chemical Society, Washington, D. C. Vol. 122, pp. 205-226, 1979. A data set may allow multiple solutions; the data set may not be definitive. Suppose that close analogs are being examined. If all the molecules have essentially the same fragments on opposite sides of a rotatable bond, then the data set is consistent with a range of solutions. Some judgment may be used in selecting a set of molecules for analysis. Judgment also may be used in the examination of the solution. Embodiments of the invention can provide a very rapid computer-aided search for a consistent solution to a very complex 3D search and matching problem. Days or weeks of analysis can be performed in minutes to a few hours, in some embodiments.
Gibbs sampling according to some embodiments of the invention can be used for more complex situations. For example, the multiple binding mode can be mimicked by applying embodiments of the invention to a data set with two sets of active compounds, D2 and D4, where each set has a different pharmacophore. Several examples were run by mixing compounds active against two different targets, as described in Xue et al., J. Chem. Inf Comput. Sci., 40, pp. 1227-1234, 2000. As was shown above, embodiments of the invention could find elements of the correct pharmacophore when the sets are analyzed separately and, again, somewhat remarkably, it can find the two pharmacophores when the two data sets are analyzed together as one data set. Of course, there is no guarantee that embodiments of the invention will always find multiple binding modes.
One potential weakness in conventional pharmacophore identification programs is the potential limitation in the number of conformations that can be used in the search process. Catalyst® appears limited to 255 conformations for each molecule. DISCO® appears to be limited to 80 conformations. Embodiments of the invention can consider only one conformation for each molecule while it is making a determination of the key features, but once it has tentative key features it can rapidly examine a large number of conformations for each molecule. Embodiments of the invention can replace a many-to-many comparison with many-to-one comparisons. For each of N molecules, compare the “model” active conformation to each of the conformations for each of the compounds under consideration. The calculations are order n rather than
the total number of conformations. It will be understood, however, that some embodiments of the invention may not check all possibilities so there is no guarantee that the process will find the optimal solution. Moreover, some embodiments of the invention may assume that there is a single binding mode (or just a few) and a consistent set of key features for the binding mode (or key features for each of the few binding modes).
A data set might allow multiple consistent solutions, so embodiments of the invention may be started multiple times and the output examined to see that if the solutions are consistent. As embodiments of the invention can be very fast, it is possible to do rigorous hold-out cross validation, as described, for example, in Burman, Biometrika, 76, pp. 503-514, 1988.
As was described in
More benchmarking may be performed to understand the range of data sets that can be successfully analyzed according to some embodiments of the invention. The computational speed can be good so larger numbers of compounds and/or larger numbers of conformations can be used. For the time being, some of the conventional dataset advice for Catalyst can be considered a guide. For example, there should be multiple, diverse compounds that appear to follow only one (or in the case of some embodiments of the invention, a few binding) mechanism. Having compounds that are fairly rigid may help. Predictors that capture potentially important information about the activity of the compounds are desirable. The compounds and their conformations can be structurally diverse.
In some embodiments of the invention, the post processing can look for and cater to multiple binding modes. Alternatively, after the execution of some embodiments of the invention, the identified key features may be used to cluster the compounds. If there are distinct clusters, then embodiments of the invention can be executed on compounds from each cluster on the reasonable assumption that these compounds from the different clusters might be binding in different ways. Note that once the key features are identified from the first run, the clustering may be more likely to correctly identify compounds binding in different ways.
Embodiments of the invention can offer a number of potential advantages for pharmacophore identification. The large datasets that can be considered at high speed because searches can be many to one. In effect, the Q vector contains the information on the key features and the features of each selected conformation are compared to it. The A matrix holds the information on the likelihood that a conformation is the binding conformation and each conformation within a molecule can be compared to A. Embodiments of the invention need not try to compare each conformation of one molecule against all the conformations of another. Relatively large data sets can be used. For example, for the 78 ACE compounds there were 46,268 conformations. The problem was run in 34 hours on a desktop Windows machine. A small data set of six compounds and 6,995 conformations was run in 36 minutes.
In the drawings and specification, there have been disclosed embodiments of the invention and, although specific terms are employed, they are used in a generic and descriptive sense only and not for purposes of limitation, the scope of the invention being set forth in the following claims.
This application claims the benefit of provisional Application No. 60/543,314, filed Feb. 10, 2004, entitled Pharmacophore Mapping, the disclosure of which is hereby incorporated herein by reference in its entirety as if set forth fully herein.
Number | Date | Country | |
---|---|---|---|
60543314 | Feb 2004 | US |