Methods for analyzing high dimensional data for classifying, diagnosing, prognosticating, and/or predicting diseases and other biological states

• US 20030207278A1
• Filed: 04/25/2002
• Published: 11/06/2003
• Est. Priority Date: 04/25/2002
• Status: Active Grant

• Current Assignee: The United States of America As Represented By The Secretary of Agriculture
• Original Assignee: The United States of America As Represented By The Secretary of Agriculture
• Inventors: Peterson, Carsten, Khan, Javed, Meltzer, Paul, Ringner, Markus
• Granted Patent: US 7,774,143 B2
• Time in Patent Office: Days
• US Class Current: 435/6
• CPC Class Codes: C12Q 1/6886 for cancer immunoassay for cancer G01N33/574
• CPC Class Codes: G16B 25/00 ICT specially adapted for hybridisation; ICT specially adapted for gene or protein expression
• CPC Class Codes: G16B 25/10 Gene or protein expression profiling; Expression-ratio estimation or normalisation
• 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 40/20 Supervised data analysis
• CPC Class Codes: G16B 5/00 ICT specially adapted for modelling or simulations in systems biology, e.g. gene-regulatory networks, protein interaction networks or metabolic networks
• CPC Class Codes: G16B 5/20 Probabilistic models

Abstract

A method of diagnosing, predicting, or prognosticating about a disease that includes obtaining experimental data, wherein the experimental data is high dimensional data, filtering the data, reducing the dimensionality of the data through use of one or more methods, training a supervised pattern recognition method, ranking individual data points from the data, wherein the ranking is dependent on the outcome of the supervised pattern recognition method, choosing multiple data points from the data, wherein the choice is based on the relative ranking of the individual data points, and using the multiple data points to determine if an unknown set of experimental data indicates a diseased condition, a predilection for a diseased condition, or a prognosis about a diseased condition.

Claims

1. A method of diagnosing, predicting, or prognosticating about a disease comprising the steps of:

(a) obtaining experimental data, wherein said experimental data comprises high dimensional data;

(b) filtering said data;

(c) reducing the dimensionality of said data through use of one or more methods;

(d) training a supervised pattern recognition method;

(e) ranking individual data points from said high dimensional data, wherein said ranking is dependent on an outcome of said supervised pattern recognition method;

(f) choosing multiple data points from said high dimensional data, wherein said choice is based on said relative ranking of said individual data points; and

(g) using said multiple data points to determine if an unknown set of experimental data indicates a diseased condition, a predilection for a diseased condition, or a prognosis about a diseased condition.

2. The method of claim 1, wherein said high dimensional data is chosen from the group consisting of:

Protein arrays, protein chips, cell array based expression analysis, analysis of patterns of single nucleotide polymorphisms in disease conditions, comparative genomic hybridization on metaphase BAC genomic, cDNA or oligonucleotide arrays and gene expression data.

3. The method of claim 2, wherein said high dimensional data is gene expression data.

4. The method of claim 3, wherein said gene expression data is obtained by using a cDNA or an oligonucleotide microarray.

6. The method of claim 1, wherein said method of reducing the dimensionality of said data is accomplished by principal component analysis.

7. The method of claim 1, wherein said supervised pattern recognition method comprises an artificial neural network.

8. The method of claim 1, wherein said multiple data points chosen from said data comprises at least 96 individual data points.

9. The method of claim 1, wherein said disease is chosen from the group consisting of multiple sclerosis, rheumatoid arthritis, and cancer.

10. The method of claim 9, wherein said disease is cancer.

11. The method of claim 10, wherein said cancer is a small round blue cell tumor.

12. The method of claim 11, wherein said cancer is neuroblastoma, rhabdomyosarcoma, non-Hodgkin lymphoma, or a Ewing family of tumors.

5. The method of 4, wherein said step of filtering said gene expression data is based on the intensity of the spots on said microarray.

13. A method of diagnosing, predicting, or prognosticating about a disease comprising the steps of:

(a) obtaining experimental data, wherein said experimental data comprises gene expression data

(b) filtering said gene expression data;

(c) reducing the dimensionality of said data through use of one or more methods;

(d) training an artificial neural network;

(e) ranking individual genes from said gene expression data, wherein said ranking is dependent on the outcome of said artificial neural network;

(f) choosing multiple genes from said gene expression data, wherein said choice is based on the relative ranking of said individual genes; and

(g) using said multiple genes to determine if an unknown set of experimental gene expression data indicates a diseased condition, a predilection for a diseased condition, or a prognosis about a diseased condition.

14. The method of claim 13, wherein said gene expression data is obtained through use of a cDNA or oligonucleotide microarray.

15. The method of claim 14, wherein said step of filtering said gene expression data is based on the intensity of the spots on said microarray.

16. The method of claim 13, wherein said step of reducing the dimensionality of said data is accomplished by principal component analysis.

17. The method of claim 13, wherein said disease is chosen from the group consisting of multiple sclerosis, rheumatoid arthritis, and cancer.

18. The method of claim 15, wherein said disease is cancer.

19. The method of claim 18, wherein said cancer is a small round blue cell tumor.

20. The method of claim 19, wherein said cancer is neuroblastoma, rhabdomyosarcoma, non-Hodgkin lymphoma, or a Ewing family of tumors.

21. A computer-based method of determining a set of multiple data points for use in diagnosing, predicting, or prognosticating about a disease, the computer-based method comprising:

(a) receiving experimental data representing high dimensional data, wherein said experimental data comprises gene expression data;

(b) filtering the experimental data;

(c) reducing the dimensionality of the experimental data using one or more methods;

(d) dividing the experimental data into a training data set and a validation data set;

(e) training an artificial neural network using the training data set to generate a trained artificial neural network;

(f) validating the performance of the trained neural network using the validation data set;

(g) generating a ranking data value corresponding to a relative ranking for individual data points using the data from the trained artificial neural network; and

(h) choosing the set of multiple data points from the data using the ranking data values.

22. The method of claim 21, wherein said gene expression data is obtained through use of a cDNA or oligonucleotide microarray.

23. The method of claim 22, wherein said step of filtering said gene expression data is based on the intensity of the spots on said microarray.

24. The method of claim 21, wherein said step of reducing the dimensionality of said data is accomplished by principal component analysis.

25. The method of claim 21, wherein said disease is chosen from the group consisting of multiple sclerosis, rheumatoid arthritis, and cancer.

26. The method of claim 23, wherein said disease is cancer.

27. The method of claim 26, wherein said cancer is a small round blue cell tumor.

28. The method of claim 26, wherein said cancer is neuroblastoma, rhabdomyosarcoma, non-Hodgkin lymphoma, or a Ewing family of tumors.

29. A computer data product readable by a computing system and encoding instructions for implementing a computer method for determining a set of multiple genes for use in diagnosing a disease, the computer-based method comprising:

receiving data representing experimental gene expression data;

filtering the gene expression data;

reducing the dimensionality of the gene expression data using principle component analysis;

dividing the gene expression data into a training data set and a validation data set;

training an artificial neural network using the training data set to generate a trained artificial neural network;

validating the performance of the trained neural network using the validation data set;

generating a ranking data value corresponding to a relative ranking for individual genes using the gene expression data in the trained artificial neural network; and

choosing the set of multiple genes from the gene expression data using the ranking data values.

30. The computer data product according to claim 29, wherein the computer data product comprises a computer readable storage medium.

31. The computer data product according to claim 29, wherein the computer data product comprises a digital data encoded within a communications carrier wave for transmission between a plurality of computing system.