Method and apparatus for designing molecules with desired properties by evolving successive populations

• US 5,434,796 A
• Filed: 06/30/1993
• Issued: 07/18/1995
• Est. Priority Date: 06/30/1993
• Status: Expired due to Term

• Current Assignee: Daylight Chemical Information Systems, Inc.
• Original Assignee: Daylight Chemical Information Systems, Inc.
• Inventors: Weininger, David
• Primary Examiner(s): Ramirez, Ellis B.
• Assistant Examiner(s): Stamber, Eric W.
• Application Number:
  US08/084,361
• Time in Patent Office: 748 Days
• Field of Search: 364/578, 364/496, 364/497, 364/498, 364/499, 436/86, 436/89, 395/13
• US Class Current: 703/12
• CPC Class Codes: B01J 19/0046 Sequential or parallel reactions, e.g. for the synthesis of polypeptides or polynucleotides; Apparatus and devices for combinatorial chemistry or for making molecular arrays synthesis methods per se C40B50/00
• CPC Class Codes: B01J 2219/007 Simulation or vitual synthesis
• CPC Class Codes: C07K 1/00 General methods for the preparation of peptides , i.e. processes for the organic chemical preparation of peptides or proteins of any length
• CPC Class Codes: C07K 1/047 Simultaneous synthesis of different peptide species; Peptide libraries
• CPC Class Codes: G16C 20/50 Molecular design, e.g. of drugs
• CPC Class Codes: G16C 20/70 Machine learning, data mining or chemometrics

Abstract

A method of and apparatus are disclosed for evolving successive populations of molecular structures and evaluating each evolved structure of each population with desired physical and/or theoretical properties. An initial population of molecules is provided in terms of representations of a number of member molecules. Evaluation is performed by a fitness function, which compares the initial population and evolved generations of member representations with the set of desired properties to provide a numerical measure or value of fitness for each structure. That numerical value indicates how closely the compared member representation corresponds with the set of desired properties. The next population is generated by changing the structure of selected molecules of a population dependent upon the numerical measure of fitness, and the process repeats. Subsequent populations evolve towards ever-better fitness. The process is terminated when an acceptable molecule evolves.

Claims

1. A method of evolving a representation of a molecule with a set of desired properties by successively generating a sequence of populations, each of said populations comprising a plurality of member representations of the chemical composition and connectivity of member molecules, said method comprising the steps of:

a) comparing each member representation of a present population of said sequence of populations with said set of desired properties to assign to each compared member representation a numerical value dependent on how closely said compared member representation corresponds with said set of desired properties;

b) selecting said member representations of said present population dependent on their respective numerical values and changing the chemical composition and connectivity of at least one of said selected member representations to be reproduced within a next population of member representations of said sequence of populations, said selecting and changing being performed on a computer; and

c) repetitively executing steps a) and b) at least one time, each execution of steps a) and b) producing said next population of said member representations within said sequence of populations.

2. The method to evolve said molecule as claimed in claim 1, wherein there is further included the step of randomly generating a first population of member representations.

3. The method to evolve said molecule as claimed in claim 2, wherein said first population of member representations is compared in accordance with step a) to produce a corresponding set of numerical values and said first population of member representations is selected and changed in accordance with step b).

4. The method to evolve said molecule as claimed in claim 1, wherein said step a) of comparing further comprises a substep of evaluating each reproduced member representation to determine whether it is chemically stable and including in said next population only those member representations which are chemically stable.

5. The method to evolve said molecule as claimed in claim 4, wherein said substep of evaluating is carried out on each member representation of said present population to produce a second numerical value indicative of its chemical stability, said numerical value and second numerical value being combined to produce a composite value, said step b) of selecting said member representations of said present population being dependent upon said composite value.

6. The method to evolve said molecule as claimed in claim 1, wherein said step b) of selecting includes selecting a number of elite member representations of said present population dependent upon their respective numerical values to be reproduced directly into said next population.

7. The method to evolve said molecule as claimed in claim 1, wherein said step b) of selecting includes a substep of selecting a number of parent member representations from said present population of member representations dependent upon their numerical values.

8. The method to evolve said molecule as claimed in claim 7, wherein said step b) of changing includes a substep of producing a selected number of child member representations from said selected number of parent member representations.

9. The method to evolve said molecule as claimed in claim 8, wherein said step b) of changing includes a substep of mutating said selected number of child member representations to provide one of said member representations to be reproduced within said next population.

10. The method to evolve said molecule as claimed in claim 9, wherein said substep of selecting selects one parent member representation, and said step b) of selecting and changing includes a substep of cloning said selected one parent member representation to reproduce a single child member representation within said next population.

11. The method to evolve said molecule as claimed in claim 9, wherein said step b) of selecting and changing further includes substeps of selecting at least two parent member representations and of breeding said two selected parent member representations to produce therefrom a single bred child member representation to be included within said next population.

12. The method to evolve said molecule as claimed in claim 11, wherein said substep of selecting includes a subsubstep of comparing said two selected parent member representations to ensure that said two selected parent member representations correspond to different member representations of said present population.

13. The method to evolve said molecule as claimed in claim 11, wherein said substep of breeding takes selected fragments of each of said two selected parent member representations, and combines said selected fragments to form said single bred child member representation to be reproduced within said next population.

14. The method to evolve said molecule as claimed in claim 11, wherein said substep of mutating includes a subsubstep of providing an atom with randomly determined properties and adding said atom to said single bred child member representation to provide one of said member representations to be reproduced within said next population.

15. The method to evolve said molecule as claimed in claim 11, wherein said substep of mutating includes a subsubstep of randomly selecting an atom of said single bred child member representation and removing it to provide one of said member representations to be reproduced within said next population.

16. The method to evolve said molecule as claimed in claim 11, where said substep of mutating includes a subsubstep of selecting an atom of said single bred child member representation and replacing it with an atom of randomly determined properties to provide one of said member representations to be reproduced within said next population.

17. The method to evolve said molecule as claimed in claim 11, wherein said substep of mutating includes a subsubstep of randomly selecting two atoms of said single bred child member representation and randomly modifying a bond between said selected two atoms to provide one of said member representations to be reproduced within said next population.

18. The method to evolve said molecule as claimed in claim 8, wherein said step b) of selecting and changing includes substeps of selecting one of said selected number of said child member representations in accordance with a finite probability and of mutating said selected one child member representation of provide one of said member representations to be reproduced within said next population.

19. The method of evolve said molecule as claimed in claim 18, wherein said substep of mutating includes a first subsubstep of selecting an atom with randomly determined properties and adding said selected atom to said selected one child member representation to provide one of said member representations to be reproduced within said next population.

20. The method to evolve said molecule as claimed in claim 19, wherein said substep of mutating includes a second subsubstep of randomly selecting an atom of said selected one child member representation and removing it to provide one of said member representations to be reproduced within said next population.

21. The method to evolve said molecule as claimed in claim 20, wherein said substep of mutating includes a third subsubstep of selecting an atom of said selected one child member representation and replacing it with an atom of randomly determined properties to provide one of said member representations to be reproduced within said next population.

22. The method to evolve said molecule as claimed in claim 21, wherein said substep of mutating includes a fourth subsubstep of randomly selecting two atoms of said selected one child member representation and randomly modifying a bond between said selected two atoms to provide one of said member representations to be reproduced with said next population.

23. The method to evolve said molecule as claimed in claim 22, wherein said substep of mutating includes a fifth subsubstep of establishing of a table for storing finite probabilities for mutating in accordance with one of said first, second, third and fourth subsubsteps of mutating each selected one child member representation produced from one or more parent member representations of said present population and a sixth subsubstep of selecting from said table for each of said selected one child member representations one of said first, second, third and fourth subsubsteps to mutate each of said selected one child member representations produced from two parent member representations of said present population to provide said member representations to be reproduced within said next population.

24. The method to evolve said molecule as claimed in claim 1, wherein said comparing step a) comprises substeps of converting said set of desired properties into a digital property representation thereof, and of converting each member representation of said present population into a digital member representation thereof and comparing each of said digital member representations one at a time with said digital property representation to determine a similarity therebetween.

25. The method of evolve said molecule as claimed in claim 1, wherein said comparing step a) comprises substeps of converting said set of desired properties into a digital fingerprint thereof, and converting each member representation of said present population into a digital structural representation thereof and comparing each of said digital structural representations one at a time with said digital fingerprint to determine a metric therebetween.

26. The method to evolve said molecule as claimed in claim 25, wherein said set of desired properties is a molecular structure of said molecule.

27. The method to evolve said molecule as claimed in claim 26, wherein said comparing step a) implements a fitness function to determine a binding energy between each of said member representations of said present population and said set of desired properties, said comparing step a) comprises substeps of constructing a model of a binding site in accordance with said set of desired properties and of carrying out a series of conformational analyses on each of said member representations of said present population to determine said binding energy between each of said member representations and said digital fingerprint.

28. The method to evolve said molecule as claimed in claim 27, wherein said binding site forms a receptor pocket, said substep of constructing said model includes generating a plurality of spheres which fill said receptor pocket and of creating a digital representation of said plurality of spheres.

29. The method to evolve said molecule as claimed in claim 28, wherein said substep of said conformational analyses randomly measures geometric distances between said plurality of spheres and a molecular structure defined by each member representation of said present population and for providing for each of said conformational analyses said numerical value in terms of said binding energy.

30. The method to evolve said molecule as claimed in claim 27, wherein each member representation of said present population comprises electrical charges associated with a plurality of atoms making up its corresponding member molecule, and said substep of constructing said model includes defining those electrical charges associated with said atoms which comprise said binding site in accordance with said digited fingerprint.

31. The method to evolve said molecule as claimed in claim 30, wherein each of said conformational analyses determines an electrostatic interaction between said electrical charges associated with one of said member representations and said electrical charges associated with said binding site to provide a corresponding numerical value.

32. The method to evolve said molecule as claimed in claim 25, wherein said set of desired properties is a representation of a class of related molecules.

33. The method to evolve said molecule as claimed in claim 32, wherein there is further included a step of analyzing each molecule of said class to determine a presence of a common characteristic of the molecules in said class and for setting corresponding bits of said member representation if said common characteristic is determined to be present.

34. The method to evolve said molecule as claimed in claim 1, wherein said step a) of comparing comprises substeps of synthesizing an actual molecule for each member representation of said present population, of introducing a sample of an actual molecule having said set of desired properties and of assaying a binding energy between each of said synthesized molecules and said molecule to provide a corresponding set of numerical values.

35. The method to evolve said molecule as claimed in claim 1, wherein said step a) of comparing performs at least one fitness function wherein a fit between said set of desired properties and each member representation of said present population is determined to provide for each member representation a single fitness value.

36. The method to evolve said molecule as claimed in claim 35, wherein said step a) of comparing performs a plurality of fitness functions on each member representation of said present population, each of said plurality of fitness functions having a corresponding set of desired properties.

37. The method to evolve said molecule as claimed in claim 36, wherein each performance of said plurality of fitness functions provides a partial function value, and said partial function values are combined for each member representation of said present population.

38. The method to evolve said molecule as claimed in claim 37, wherein said partial function value for each fitness function is scaled to express said partial function values provided from said plurality of fitness functions in common units.

39. The method to evolve said molecule as claimed in claim 1, wherein said step b) of selecting provides member representations of said next population only from selected member representations of said present population.

40. The method to evolve said molecule as claimed in claim 1, wherein each population of said sequence of populations comprises a given number of said member representations.