Artificial intelligence and global normalization methods for genotyping

• US 7,035,740 B2
• Filed: 03/24/2004
• Issued: 04/25/2006
• Est. Priority Date: 03/24/2004
• Status: Active Grant

• Current Assignee: Illumina Incorporated
• Original Assignee: Illumina Incorporated
• Inventors: Kermani, Bahram Ghaffarzadeh
• Primary Examiner(s): Brusca, John
• Assistant Examiner(s): Zhou, Shubo (Joe)
• Application Number:
  US10/809,608
• Publication Number:
  US 20050216207A1
• Time in Patent Office: 762 Days
• Field of Search: 435/4, 435/6, 536/23.1, 536/24.3, 702/19, 702/20
• US Class Current: 702/19
• CPC Class Codes: G16B 20/00 ICT specially adapted for functional genomics or proteomics, e.g. genotype-phenotype associations
• CPC Class Codes: G16B 20/20 Allele or variant detection, e.g. single nucleotide polymorphism [SNP] detection
• CPC Class Codes: G16B 20/40 Population genetics; Linkage disequilibrium
• CPC Class Codes: G16B 40/00 ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding
• CPC Class Codes: G16B 40/10 Signal processing, e.g. from mass spectrometry [MS] or from PCR
• CPC Class Codes: G16B 40/20 Supervised data analysis
• CPC Class Codes: G16B 40/30 Unsupervised data analysis

Abstract

Described herein are systems and methods for normalizing data without the use of external controls. Also described herein are systems and methods for analyzing cluster data, such as genotyping data, using an artificial neural network.

Claims

1. A method of normalizing genetic data for n loci, wherein n is an integer greater than one, comprising

(a) obtaining genetic data comprising n sets of first and second signal values related in a coordinate system, wherein said first and second signal values are indicative of the levels of a first and second allele, respectively, at n loci;

(b) identifying a set of sweep points in said coordinate system;

(c) identifying a set of control points, said control points comprising at least a subset of said signal values that are proximal to said sweep points;

(d) projecting said control points to a line or curve passing through said sweep points, thereby forming set points;

(e) determining parameters of a registration transformation equation based on said set of control points and said set points; and

(f) transforming said n sets of first and second signal values according to said registration transformation equation and said parameters, thereby normalizing said genetic data.

2. The method of claim 1, wherein said genetic data is represented in a graphical format.

3. The method of claim 2, wherein said graphical format comprises Cartesian coordinates.

4. The method of claim 1, wherein said genetic data is provided in a tabular format.

5. The method of claim 1, wherein said identifying sweep points comprises

(i) identifying an upper limit on a line or curve through said coordinate system;

and

(ii) locating said sweep points between the origin of each axis and said upper limit.

6. The method of claim 5, wherein said upper limit has a value in a first dimension that is greater than or equal to the first dimension of any of said signal values.

7. The method of claim 5, further comprising a step of identifying a lower limit on said line or curve, and wherein said locating comprises locating said sweep points between said lower limit and said upper limit.

8. The method of claim 1, wherein said identifying a set of control points comprises triangulation using pairs of signal values and a sweep point.

9. The method of claim 8, wherein said triangulation comprises Delaunay triangulation.

10. The method of claim 1, wherein said identifying a set of control points comprises computing all pair-wise distances between the signal values and each sweep point.

11. The method of claim 1, wherein said registration transformation equation comprises affine transformation projecting said control points onto said set points.

12. The method of claim 1, wherein said registration transformation equation comprises linear conformational transformation projecting said control points onto said set points.

13. The method of claim 1, wherein said registration transformation equation comprises projective transformation projecting said control points onto said set points.

14. The method of claim 1, wherein said registration transformation equation comprises polynomial transformation projecting said control points onto said set points.

15. The method of claim 1, wherein said determining parameters of a registration transformation equation comprises global registration.

16. The method of claim 1, wherein said set of control points is fewer in number compared to the number of first and second signal values.

17. The method of claim 1, wherein said sweep points are located on a line or curve through said coordinate system when represented graphically.

18. The method of claim 17, wherein said line comprises an axis of said coordinate system.

19. The method of claim 1, wherein said sweep points are spaced along said line or curve in a manner selected from the group consisting of linear, log-linear and non-linear.

20. The method of claim 1, wherein said coordinate system comprises two dimensions.

21. The method of claim 20, wherein step (b) comprises identifying two sets of sweep points in said coordinate system;

and step (c) comprises identifying two sets of control points.

22. The method of claim 1, wherein said genetic data comprises n sets of first, second and third signal values related in a coordinate system, wherein said first, second and third signal values are indicative of the levels of a first, second and third allele, respectively, at n loci.

23. The method of claim 22, wherein said coordinate system comprises three dimensions.

24. The method of claim 22, wherein step (b) comprises identifying three sets of sweep paints in said coordinate system;

and step (c) comprises identifying three sets of control points.

25. The method of claim 1, wherein said registration transformation is selected from the group consisting of rotation of said it sets of fast and second signal values, translation of said n sets of first and second signal values, scaling of said n sets of first and second signal values, and sheer of said n sets of first and second signal values.

26. The method of claim 1, further comprising a step of balancing said n sets of first and second signal values by a balancing signal transformation, thereby balancing the probability function for the distribution of said n sets of first and second signal values as a function of signal intensity.

27. The method of claim 1, wherein said balancing signal transformation is selected from the group consisting of natural logarithm, base 2 logarithm, base 10 logarithm, arctangent, square root, nth root, wherein n>

2, and Box-Cox.

28. The method of claim 1, wherein step (a) comprises obtaining genetic data comprising n sets of first and second signal values for n loci related in a coordinate system, wherein said first signal value is indicative of the level of a first allele at a locus of said n loci and said second signal value is indicative of the level of a second allele at said locus.

29. A system for analysis of genetic data, comprising

(a) an array reader configured to detect signals from separate locations on an array substrate;

(b) a computer processor configured to receive signal values from said array reader;

(c) a normalization module comprising commands for

(i) reading said signal values;

(ii) identifying a set of sweep points for said signal values in a coordinate system;

(iii) identifying a set of control points, said control points comprising at least a subset of said signal values that are proximal to said sweep points;

(iv) projecting said control points to a line or curve passing through said sweep points, thereby forming set points;

(v) determining parameters of a registration transformation equation based on said control points and said set points; and

(vi) transforming said signal values according to said registration transformation equation and said parameters, thereby providing normalized genetic data; and

(d) a clustering module comprising commands for

(i) reading said normalized genetic data;

(ii) comparing fit of said normalized genetic data to each of a plurality of cluster models using an artificial neural network, thereby determining a best fit cluster model; and

(iii) assigning said signal values to at least one cluster according to said best fit cluster model, wherein if said best fit cluster model contains at least one actual cluster and at least one missing cluster, then using a second artificial neural network to propose a location for said at least one missing cluster.

32. The system of claim 29, wherein said commands for identifying sweep points comprise commands for

(i) identifying an upper limit on a line or curve through said coordinate system;

and

(ii) locating said sweep points between the origin of each axis and said upper limit.

33. The system of claim 32, wherein said normalization module further comprises commands for identifying a lower limit on said line or curve, and wherein said locating comprises locating said sweep points between said lower limit and said upper limit.

34. The system of claim 29, wherein said identifying a set of control points comprises triangulation using pairs of signal values and a sweep point.

35. The system of claim 34, wherein said triangulation comprises Delaunay triangulation.

36. The system of claim 29, wherein said identifying a set of control points comprises computing all pair-wise distances between the signal values and each sweep point.

37. The system of claim 29, wherein said registration transformation equation comprises affine transformation projecting said control points onto said set points.

38. The system of claim 29, wherein said registration transformation equation comprises linear conformational transformation projecting said control points onto said set points.

39. The system of claim 29, wherein said registration transformation equation comprises projective transformation projecting said control points onto said set points.

40. The system of claim 29, wherein said registration transformation equation comprises polynomial transformation projecting said control points onto said set points.

41. The system of claim 29, wherein said determining parameters of a registration transformation equation comprises global registration.

42. The system of claim 29, wherein said sweep points are located on a line or curve through said coordinate system when represented graphically.

43. The system of claim 42, wherein said line comprises an axis of said coordinate system.

44. The system of claim 29, wherein said sweep points are spaced along said line or curve in a manner selected from the group consisting of linear, log-linear and non-linear.

45. The system of claim 29, wherein said coordinate system comprises two dimensions.

46. The system of claim 29, wherein said coordinate system comprises three dimensions.

30. A method of determining the alleles present at n loci for an individual, comprising

(a) obtaining genetic data comprising n sets of first and second signal values related in a coordinate system, wherein said first and second signal values are indicative of the levels of a first and second allele, respectively, at n loci;

(b) identifying a set of sweep points in said coordinate system;

(c) identifying a set of control points, said control points comprising at least a subset of said signal values that are proximal to said sweep points;

(d) projecting said control points to a line or curve passing through said sweep points, thereby forming set points;

(e) determining parameters of a registration transformation equation based on said set of control points and said set points; and

(f) transforming said n sets of first and second signal values according to said registration transformation equation and said parameters, thereby normalizing said genetic data;

(g) comparing fit of said normalized genetic data to each of a plurality of cluster models using an artificial neural network, thereby determining a best fit cluster model;

(h) assigning said signal values to at least one cluster according to said best fit cluster model, wherein if said best fit cluster model contains at least one actual cluster and at least one missing cluster, then using a second artificial neural network to propose a location for said at least one missing cluster; and

(i) determining, for an individual, the alleles present at said n loci.

31. The method of claim 30, wherein step (a) comprises obtaining genetic data comprising n sets of first and second signal values for n loci related in a coordinate system, wherein said first signal value is indicative of the level of a first allele at a locus of said n loci and said second signal value is indicative of the level of a second allele at said locus.

47. The method of claim 30, wherein said genetic data is represented in a graphical format.

48. The method of claim 47, wherein said graphical format comprises Cartesian coordinates.

49. The method of claim 30, wherein said genetic data is provided in a tabular format.

50. The method of claim 30, wherein said identifying sweep points comprises

(i) identifying an upper limit on a line or curve through said coordinate system;

and

(ii) locating said sweep points between the origin of each axis and said upper limit.

51. The method of claim 50, wherein said upper limit has a value in a first dimension that is greater than or equal to the first dimension of any of said signal values.

52. The method of claim 50, further comprising a step of identifying a lower limit on said line or curve, and wherein said locating comprises locating said sweep points between said lower limit and said upper limit.

53. The method of claim 30, wherein said identifying a set of control points comprises triangulation using pairs of signal values and a sweep point.

54. The method of claim 53, wherein said triangulation comprises Delaunay triangulation.

55. The method of claim 30, wherein said identifying a set of control points comprises computing all pair-wise distances between the signal values and each sweep point.

56. The method of claim 30, wherein said registration transformation equation comprises affine transformation projecting said control points onto said set points.

57. The method of claim 30, wherein said registration transformation equation comprises linear conformational transformation projecting said control points onto said set points.

58. The method of claim 30, wherein said registration transformation equation comprises projective transformation projecting said control points onto said set points.

59. The method of claim 30, wherein said registration transformation equation comprises polynomial transformation projecting said control points onto said set points.

60. The method of claim 30, wherein said determining parameters of a registration transformation equation comprises global registration.

61. The method of claim 30, wherein said set of control points is fewer in number compared to the number of first and second signal values.

62. The method of claim 30, wherein said sweep points are located on a line or curve through said coordinate system when represented graphically.

63. The method of claim 62, wherein said line comprises an axis of said coordinate system.

64. The method of claim 30, wherein said sweep points are spaced along said line or curve in a manner selected from the group consisting of linear, log-linear and non-linear.

65. The method of claim 30, wherein said coordinate system comprises two dimensions.

66. The method of claim 65, wherein step (b) comprises identifying two sets of sweep points in said coordinate system;

and step (c) comprises identifying two sets of control points.

67. The method of claim 30, wherein said genetic data comprises n sets of first, second and third signal values related in a coordinate system, wherein said first, second and third signal values are indicative of the levels of a first, second and third allele, respectively, at n loci.

68. The method of claim 67, wherein said coordinate system comprises three dimensions.

69. The method of claim 67, wherein step (b) comprises identifying three sets of sweep points in said coordinate system;

and step (c) comprises identifying three sets of control points.

70. The method of claim 30, wherein said registration transformation is selected from the group consisting of rotation of said n sets of first and second signal values, translation of said n sets of first and second signal values, scaling of said n sets of first and second signal values, and sheer of said n sets of first and second signal values.

71. The method of claim 30, further comprising a step of balancing paid n sets of first and second signal values by a signal transformation, thereby balancing the probability function for the distribution of said a sets of first and second signal values as a function of signal intensity.

72. The method of claim 30, wherein said signal transformation is selected from the group consisting of natural logarithm, base 2 logarithm, base 10 logarithm, arctangent, square root, nth root wherein n>

2, and Box-Cox.