Method and system for artificial intelligence directed lead discovery though multi-domain agglomerative clustering

A computer-operable method for identifying pharmacophoric mechanisms through analysis of a plurality of molecules, each molecule having a respective feature characteristic and a respective activity characteristic, the method comprising, in combination:

• (a) establishing in a computer memory a plurality of cluster objects, each cluster object representing one of the molecules;
• (b) agglomeratively clustering the cluster objects based on comparisons of the feature characteristics of the molecules that the cluster objects represent, to build in the computer memory a hierarchical pyramid comprising a plurality of cluster objects, each cluster object of the pyramid representing a number of the molecules;
• (c) with respect to each cluster object of a plurality of the cluster objects of the pyramid, identifying a substructure common to molecules represented the cluster object, each substructure defining a respective pharmacophoric mechanism;
• and
• (d) outputting for viewing by a person a description of at least a portion of the hierarchical pyramid, including at least one substructure identified in step (c).

A computer-operable method for identifying pharmacophoric mechanisms through analysis of a plurality of molecules, each molecule defining a feature characteristic and an activity characteristic, the method comprising the following steps:

• storing in a computer memory a plurality of data objects, each data object representing one of the molecules and having associated therewith a feature vector representing the feature characteristic of the molecule;
• pyramidally clustering the data objects based on their associated feature vectors, to form in the computer memory a pyramidal data structure comprising a number of nodes, each node representing one or more of the molecules;
• when pyramidally clustering the data objects, encountering a tie in proximity between a given node and at least two other nodes and responsively merging the given node separately with each of the at least two other nodes;
• with respect to each node of the pyramidal data structure, identifying a chemical feature set common to the one or more molecules represented by the node, the chemical feature set defining a pharmacophore;
• and
• providing an output describing at least a portion of the pyramidal data structure and including a description of the chemical feature set identified with respect to at least one node of the pyramidal data structure.

A computer-readable medium embodying a set of machine language instructions executable by a computer to identify pharmacophoric mechanisms through analysis of a plurality of molecules, each molecule defining a feature characteristic and an activity characteristic, wherein the machine language instructions are executable by the computer to perform functions comprising:

• storing in a computer memory a plurality of data objects, each data object representing one of the molecules and having associated therewith a feature vector representing the feature characteristic of the molecule;
• pyramidally clustering the data objects based on their associated feature vectors, to form in the computer memory a pyramidal data structure comprising a number of nodes, each node representing one or more of the molecules,
• when pyramidally clustering the data objects, encountering a tie in proximity between a given node and at least two other nodes and responsively merging the given node separately with each of the at least two other nodes;
• with respect to each node of the pyramidal data structure, identifying a chemical feature set common to the one or more molecules represented by the node, the chemical feature set defining a pharmacophore;
• and
• providing an output describing at least a portion of the pyramidal data structure and including a description of the chemical feature set identified with respect to at least one node of the pyramidal data structure.

A computer-operable method for learning pharmacophoric mechanisms through analysis of a plurality of molecules, each molecule having a respective feature characteristic and a respective activity characteristic, the method comprising, in combination:

• (a) selecting from the plurality of molecules a group of molecules having at least a threshold activity characteristic;
• (b) storing in a data storage medium a plurality of data objects each representing at least one of the molecules of the group, at least a first data object representing a plurality of molecules;
• (c) measuring distances between the data objects based on the feature characteristics of the molecules represented by the data objects, and making a determination that the distance between the first data object and a second data object is substantially equal to the distance between the first data object and a third data object;
• (d) in response to the determination, (i) storing in the data storage medium a fourth data object representing the molecules cooperatively represented by the first data object and the second data object and (ii) storing in the data storage medium a fifth data object representing the molecules cooperatively represented by the first data object and the third data object;
• (e) identifying at least (i) a common feature set among the feature characteristics of the molecules represented by the first data object and (ii) a common feature set among the feature characteristics of the molecules represented by the fourth data object, whereby each common feature set defines a respective pharmacophoric mechanism;
• and
• (f) providing to a person an indication of at least the common feature sets identified with respect to the molecules of the first and fourth data objects.

A computer-operable method for analyzing a plurality of molecules, each molecule having a respective feature characteristic and a respective activity characteristic, wherein the respective activity characteristic of each molecule represents at least a threshold activity level, the method comprising, in combination:

• (a) storing in a computer memory a plurality of cluster objects, each cluster object representing at least one of the molecules;
• (b) conducting a merging process with respect to the cluster objects, the merging process comprising:
• (i) comparing pairs of the cluster objects and, for each pair, measuring a respective dissimilarity between the cluster objects within the pair based on the feature characteristics of the molecules represented by the respective cluster objects;
• (ii) of the dissimilarities measured in step (i), identifying a smallest dissimilarity;
• (iii) selecting at least one pair of the cluster objects that has the smallest measured dissimilarity;
• and
• (iv) with respect to each of the at least one pair selected in step (iii), merging the cluster objects of the pair to establish a cluster object cooperatively representing the molecules that were represented by the cluster objects of the pair;
• (c) if at least two cluster objects have not yet been merged, then repeating step (b) with respect to the cluster objects that have not yet been merged;
• (d) with respect to at least each cluster object established in step (b)(iv), identifying a common substructure among the molecules represented by the cluster object;
• and
• (e) outputting a description of at least one cluster object established in step (b)(iv), wherein, the description of each of the at least one cluster object comprises a first portion indicating the common substructure identified in step (d) for the cluster object.

A computer-readable medium embodying a set of machine language instructions executable by a computer to analyze a plurality of molecules, each molecule having a respective feature characteristic and a respective activity characteristic, wherein the respective activity characteristic of each molecule represents at least a threshold activity level, wherein the machine language instructions are executable by the computer to perform functions comprising:

• (a) storing in a computer memory a plurality of cluster objects, each cluster object representing at least one of the molecules,
• (b) conducting a merging process with respect to the cluster objects, the merging process comprising:
• (i) comparing pairs of the cluster objects and for each pair, measuring a respective dissimilarity between the cluster objects within the pair based on the feature characteristics of the molecules represented by the respective cluster objects;
• (ii) of the dissimilarities measured in step (i), identifying a smallest dissimilarity,
• (iii) selecting at least one pair of the cluster objects that has the smallest measured dissimilarity;
• and
• (iv) with respect to each of the at least one pair selected in step (iii), merging the cluster objects of the pair to establish a cluster object cooperatively representing the molecules that were represented by the cluster objects of the pair;
• (c) if at least two cluster objects have not yet been merged, then repeating step (b) with respect to the cluster objects that have not yet been merged;
• (d) with respect to at least each cluster object established in step (b)(iv), identifying a common substructure among the molecules represented by the cluster object;
• and
• (e) outputting a description of at least one cluster object established in step (b)(iv), wherein, the description of each of the at least one cluster object comprises a first portion indicating the common substructure identified in step (d) for the cluster object.

A processing system for screening a data set representing a plurality of molecules, in order to assist in identifying sets of molecular features that are likely to correlate with specified activity, the data set defining, for each represented molecule, a feature characteristic and an activity characteristic, the processing system comprising, in combination:

• at least one processor;
• at least one data storage medium;
• machine-language instructions stored in the at least one data storage medium and executable by the at least one processor to perform the following functions:
• (a) storing in a computer memory a plurality of cluster objects, each cluster object representing one of the molecules;
• (b) with respect to the cluster objects, conducting a merging process comprising:
• (i) comparing pairs of the cluster objects and, for each pair, measuring a respective dissimilarity between the cluster objects within the pair, based on the feature characteristics of the molecules represented by the cluster objects of the pair;
• (ii) of the dissimilarities measured in step (i), identifying a smallest dissimilarity;
• (iii) selecting at least one pair of the cluster objects that has the smallest respective measured dissimilarity;
• (iv) with respect to each pair of the at least one pair selected in step (iii), merging the cluster objects of the pair to establish a cluster object cooperatively representing the molecules that were represented by the cluster objects of the pair;
• (c) if at least two cluster objects have not yet been merged, then repeating step (b) with respect to the cluster objects that have not yet been merged;
• (d) with respect to at least each cluster object established in step (b)(iv) identifying a common substructure among the molecules represented by the cluster object;
• and
• (e) outputting a description of at least one cluster object established in step (b)(iv), wherein, the description of each of the at least one cluster object comprises a first portion indicating the common substructure identified in step (d) for the cluster object.