Method and system for robust classification strategy for cancer detection from mass spectrometry data

• US 20080025591A1
• Filed: 07/27/2006
• Published: 01/31/2008
• Est. Priority Date: 07/27/2006
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Levine, Arnold J., Bhanot, Gyan Vatsaraja, Alexe, Gabriele, Venkataraghavan, Babu
• Granted Patent: US 7,899,625 B2
• Time in Patent Office: Days
• US Class Current: 382/132
• CPC Class Codes: G06F 18/217 Validation; Performance evaluation; Active pattern learning techniques
• CPC Class Codes: G06F 18/40 Software arrangements specially adapted for pattern recognition, e.g. user interfaces or toolboxes therefor
• CPC Class Codes: G06F 2218/12 Classification; Matching

Abstract

A robust classification method for cancer detection from mass spectrometry data includes inputting the mass spectrometry data, preprocessing the spectrometry data, conducting robust feature selection, generating predictions for the test data sets using multiple data classifiers, the multiple data classifiers including artificial neural networks, support vector machines, weighted voting on data patterns, classification and regression trees, k-nearest neighbor classification, and logistic regression, and constructing and validating a meta-classifier by combining individual predictions of the multiple data classifiers to generate a robust prediction of a phenotype. The test data sets are used exclusively for validation of the meta-classifier.

Claims

1. A robust classification method for cancer detection from mass spectrometry data, comprising:

preprocessing mass spectrometry data;

conducting robust feature selection from the mass spectrometry data;

generating predictions for test data sets using multiple data classifiers, said multiple data classifiers comprising artificial neural networks, support vector machines, weighted voting on data patterns, classification and regression trees, k-nearest neighbor classification, and logistic regression; and

constructing and validating a meta-classifier by combining and averaging individual predictions of said multiple data classifiers to generate a robust prediction of a phenotype,

wherein said test data sets are used exclusively for validation of the meta-classifier.

2. The method according to claim 1, further comprising:

mapping said individual classifiers to generate intermediate classifiers.

3. A classification method, comprising:

applying machine learning tools on a set of data to create a robust predictive model.

4. The classification method according to claim 3, wherein said predictive model detects a presence of cancer.

5. The classification method according to claim 3, wherein said set of data comprises mass spectrometry data.

6. The classification method according to claim 3, further comprising:

conducting robust feature selection of said set of data;

generating predictions for said set of data; and

validating said predictions.

7. The classification method according to claim 6, further comprising:

inputting said set of data, prior to said conducting robust feature selection.

8. The classification method according to claim 6, further comprising:

preprocessing said set of data, prior to conducting said robust feature selection.

9. The classification method according to claim 6, wherein said robust feature selection comprises:

peak extraction to identify peaks in a data spectrum and to extract peaks from background data in the data spectrum.

10. The classification method according to claim 9, wherein said robust feature selection comprises:

filtering data peaks extracted by said peak extraction.

11. The classification method according to claim 6, wherein said robust feature selection comprises:

selecting a support set of data on which an accurate weighting pattern-based model can be constructed by using a combinatorial pattern recognition algorithm to extract a large collection of data patterns from the set of data and from perturbations with experimental noise of the set of data.

12. The classification method according to claim 6, wherein said generating predictions for said set of data comprises generating prediction using multiple data classifiers.

13. The classification method according to claim 6, wherein said validating said predictions comprises constructing and validating a meta-classifier by combining and averaging individual predictions of a plurality of data classifiers to generate a robust prediction.

14. The classification method according to claim 8, wherein said preprocessing said set of data comprises:

creating a training data set and test data sets.

15. The classification method according to claim 14, wherein said test data sets are used for said validating said predictions.

16. The classification method according to claim 12, further comprising:

mapping said multiple data classifiers to generate intermediate classifiers.

17. The classification method according to claim 16, further comprising:

linearly combining said intermediate classifiers to generate a meta-classifier.

19. A computer-readable medium tangibly embodying a program of computer readable instructions executable by a digital processing apparatus to perform the classification method in accordance with claim 3.

20. A method of deploying computing infrastructure, comprising integrating computer-readable code into a computing system, wherein the computer readable code in combination with the computing system is capable of performing the classification method in accordance with claim 3.

18. A classification system, comprising:

a feature selection unit that conducts robust feature selection of a set of data;

a prediction unit that generates predictions for said set of data; and

a validating unit that validates said predictions.