US 6,240,374 B1 · Further method of creating and rapidly searching a virtual library of potential molecules using validated molecular structural descriptors

Abstract

The problem of how to select out of a large chemically accessible universe molecules representative of the diversity of that universe is resolved by the discovery of a method to validate molecular structural descriptors. Using the validated descriptors, optimally diverse subsets can be selected. In addition, from the universe, molecules with characteristics similar to a selected molecule can be identified. The validated descriptors also enable the generation of a huge virtual library of potential product molecules which could be formed by combinatorial arrangement of structural variations and cores. In this virtual library it is possible to search billions of possible product compounds in relatively short time frames.

Claims38

Dependent claims

A computer-based method for generating a virtual library of component parts and their characteristics in which all possible product molecules combinatorially derived from the component parts can be searched, without the necessity of generating the product structures during the search, for product molecules having desired properties by searching through only a combination of the descriptore of the component parts of the product molecules comprising the following steps:

a. defining chemical transformations and reagents and cores to be used to specify possible product molecules;
and
b. using appropriate molecular descriptors, validated as possessing a neighborhood property, to precalculate characteristics of the component parts of all possible product molecules.

A computer-based method for generating and searching a virtual library of structural variations, cores, and their associated molecular structural descriptors, which can be searched for product molecules, derived from the combinatorial assembly of the structural variations and cores, having a high probability of sharing an activity possessed by a molecule of interest not known to be derived from a combinatorial reaction if the descriptor values of the molecules are within the neighborhood distance of the descriptor values of the molecule of interest, without the necessity of generating the product structures during the search, comprising the following steps:

a. generating a virtual library by:
(1). creating one or more files identifying one or more combinatorial reactions for one or more core structures;
(2). creating separate structural variation files, associated with the reaction identifying files, in which are listed together the structural variations representative of those reactants which will react at each variation site of each combinatorial reaction;
(3). associating with each structural variation, data, characterizing each structural variation including:
(a) characterizing data, which has not been derived from applying at least one molecular structural descriptor, validated as possessing a neighborhood property, to the structural variations, taking into account the structures of the cores with which the structural variations would be combined in the combinatorial syntheses;
and
(b) characterizing data which has been derived from applying at least one molecular structural descriptor, validated as possessing a neighborhood property, to the structural variations taking into account to the extent appropriate for application of the descriptor the structures of the cores with which the structural variations would be combined in the combinatorial syntheses;
(4). associating with each core, data, characterizing each core including:
(a). characterization data which has not been derived from application of molecular descriptors, validated as possessing a neighborhood property;
and
(b). characterizing data which is derived from the following additional steps:
(i). selecting a first core;
(ii). selecting an attachment bond on the core;
(iii). topomerically aligning the core;
(iv). characterizing the core with CoMFA fields and the coordinates of the end points of the other attachment bonds;
(v). repeating steps (ii) through (iv) for all attachment bonds on the core;
(vi). selecting a next core;
and
(vii). repeating steps (ii) through (vi) for all cores;
b. fragmenting the molecule of interest as described in a fragmentation table;
c. selecting a fragmentation pattern;
d. aligning the fragments according to topomeric alignment rules;
e. generating CoMFA fields for each aligned fragment;
f. identifing whether the fragmentation pattern generated a core, and, if so, implementing the following steps:
(1). selecting an attachment bond on the core;
(2). topomerically aligning the core;
(3). characterizing the core with CoMFA fields and coordinates of end points of the other attachment bonds;
(4). repeating steps (1) through (3) for all attachment bonds on the core;
(5). calculating the differences in CoMFA fields and differences in attachment bond coordinates for all attachment bond orientations;
g. identifying which reaction types within the virtual library correspond to the reaction type resulting from the fragmentation;
h. selecting any of the corresponding reaction types in the virtual library identified in step g;
i. if the fragmentation pattern generated a core, determining acceptable cores from the reaction type in the virtual library selected in step h by the following steps:
(1). selecting any core from the virtual library;
(2). generating a list of all the possible ways the attachment bonds in the virtual library core correspond to the attachment bonds in the fragmentation core by the following steps:
(a). selecting a corresponding way;
(b). determining the root sum of squared differences in attachment bond coordinates over all attachment bonds and of CoMFA field values;
(c). if the difference determined in step (b) is greater than a chosen neighborhood value, delete the corresponding way from the list;
(d). repeat steps (a) through (c) for all corresponding ways;
(3). repeat steps (1) through (2) until a core is found with at least one corresponding way;
j. selecting the structural variations matching the reaction type in step h which were used in generating the virtual library;
k. selecting and outputting the combinations of structural variations which yield a product molecule falling within a chosen neighborhood value by the following steps:
(1). selecting any combination of structural variations;
(2). selecting a way in which the combination of structural variations may correspond to the fragments generated in step d;
(3). if the fragmentation pattern generated a core and the correspondence of step (2) is not on the list generated in step i, proceed to step k.(7);
(4). determining the root sum of squared differences between the CoMFA fields for the selected correspondence and combination of structural variations and the fields generated in step e;
(5). if the fragmentation pattern generated a core, adding to the root sum determined in step k.(4) the root sum determined in step i.(2)(b);
(6). if the total of the root sums determined in step (5) is less than a chosen neighborhood value, output the selected combination of structural variations, with the core, if any, and proceed to step k.(8);
(7). repeat steps k.(2) through k.(6) until either the combination of structural variations has been output in step k.(6) or until no more possible correspondences remain;
l. if the fragmentation pattern generated a core, repeat steps i through k until all cores in the matching reaction type of the virtual library have been considered;
m. repeat steps h through l until all matching reaction types in the virtual library have been considered;
n. repeat steps c through m until all fragmentation patterns in the table have been considered.

48 more elements

A computer-based method for identifying from a virtual library of structural variations, cores, and their associated molecular structural descriptors, which can be searched for product molecules derived from the combinatorial assembly of the structural variations and cores, those molecules having a high probability of sharing an activity possessed by a molecule of interest of unknown synthetic pathway, if the descriptor values of the molecules are within the neighborhood distance of the descriptor values of the molecule of interest, comprising the following steps:

a. fragmenting the query molecule as described in a fragmentation table;
b. by using molecular structural descriptors, validated as possessing a neighborhood property, with which the virtual library was generated, generating the descriptor values for each fragment;
and
c. searching the virtual library for structural variations whose descriptor values fall within the neighborhood distance of the descriptor values for each fragment and identifying all possible combinatorial product molecules which could result from the specified structural variations.

1 more element

A computer-based method for generating a virtual library of structural variations, cores, and their associated molecular structural descriptors, which can be searched for product molecules derived from the combinatorial assembly of the structural variations and cores having desired properties, by combining descriptors of the structural variations and cores to generate descriptors representative of the product molecules, without the necessity of generating the product structures during the search, comprising the following steps:

a. creating one or more files identifying one or more combinatorial reactions for one or more core structures;
b. creating separate structural variation files, associated with the reaction identifying files, in which are listed together the structural variations representative of those reactants which will react at each variation site of each combinatorial reaction;
c. associating with each structural variation, data, characterizing each structural variation including:
(1) characterizing data, which has not been derived from applying at least one molecular structural descriptor, validated as possessing a neighborhood property, to the structural variations, taking into account the structures of the cores with which the structural variations would be combined in the combinatorial syntheses;
and
(2) characterizing data which has been derived from applying at least one molecular structural descriptor, validated as possessing a neighborhood property, to the structural variations taking into account to the extent appropriate for application of the descriptor the structures of the cores with which the structural variations would be combined in the combinatorial syntheses.

3 more elements

Dependencies are inferred from the claim text.

Further method of creating and rapidly searching a virtual library of potential molecules using validated molecular structural descriptors

Life of this patent

Further method of creating and rapidly searching a virtual library of potential molecules using validated molecular structural descriptors

Life of this patent

Abstract

Claims38