Method of non-linear analysis of biological sequence data

• US 20030097227A1
• Filed: 03/27/2002
• Published: 05/22/2003
• Est. Priority Date: 03/29/2001
• Status: Active Grant

• Current Assignee: E. I. du Pont de Nemours and Company (Corteva, Inc.)
• Original Assignee: E. I. du Pont de Nemours and Company (Corteva, Inc.)
• Inventors: Arce, Gonzalo Ramiro, Bloch, Karen Marie
• Granted Patent: US 6,728,642 B2
• Time in Patent Office: Days
• US Class Current: 702/75
• CPC Class Codes: G16B 20/00 ICT specially adapted for functional genomics or proteomics, e.g. genotype-phenotype associations
• 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 40/30 Unsupervised data analysis

Abstract

A method of classifying biological elements into functional families includes the steps of representing a characteristic numerically, performing a time-frequency transform on the numeric representation, and identifying biological elements of a common functional family by clusters of data having a common frequency characteristic in the time-frequency domain. Biologically active regions on those biological elements are identified by an amplitude analysis of the clusters.

Claims

1. A method of classifying a biological element comprised of biological subelements into a functional family, wherein each family is represented by a cluster of data points around a common frequency characteristic of a time-frequency transform, the method comprising the steps of:

a) converting a symbolic representation of a sequence of biological subelements of the biological element to a numeric representation of that sequence;

b) performing a time-frequency transform on the numeric representation;

c) identifying a cluster of data having a common frequency characteristic in the time-frequency domain,

thereby to identify the biological element in the functional family corresponding to that cluster of data points.

2. The method of claim 1 further comprising, after step b), the step of:

b1) filtering interference terms from the transformed data.

3. The method of claim 2 wherein, the time-frequency transform is the Wigner-Ville time-frequency transform.

4. The method of claim 3 wherein, the interference terms of the Wigner-Ville time-frequency transform are filtered using the center affine filter method.

5. The method of claim 1 wherein the numeric representation is a scalar representation of a functional characteristic of each biological subelement of the biological element.

6. The method of claim 1 wherein the numeric representation is a vector representation of multiple functional characteristics of each subelement of the biological element.

7. The method of claim 6, wherein the vector representation is reduced to a minimal set of dimensions which preserves the functionally important characteristics of one or more subelements of the biological element.

8. The method of claim 1, further comprising the step of:

d) identifying clusters of data points whose amplitude exceeds a predetermined threshold, thereby to identify biologically active regions in the biological element.

9. The method of claim 1, further comprising the step of:

d) repeating steps a) through c) for a second biological element, and

e) identifying clusters of data points whose amplitude exceeds a predetermined threshold,

thereby to identify relationships among the first and second biological elements.

10. The method of claim 1 further comprising the step of:

after step b), plotting the resulting transformed data in the time-frequency domain to create a visual representation of the transformed data.

11. A method of classifying proteins into functional families comprising the steps of:

a) converting a symbolic representation of a primary amino acid sequence data to a numeric representation of that sequence;

b) performing a time-frequency transform on the numeric representation;

c) identifying clusters of data having a common frequency characteristic in the time-frequency domain, thereby to identify proteins of a common functional family.

12. The method of claim 11 wherein the numeric representation is a scalar representation of a physico-chemical property of each amino acid of the primary amino acid sequence data.

13. The method of claim 11, wherein the numeric representation is a vector representation of multiple physico-chemical properties.

14. The method of claim 13, wherein the vector representation is reduced to a minimal set of dimensions which preserves the functionally important features of the protein.

15. The method of claim 11, further comprising the step of:

d) identifying clusters of data points whose amplitude exceeds a predetermined threshold,

thereby to identify biologically active regions in the protein.

16. A method of classifying a gene in an organism into a functional family from a set of expression data for a set of genes, the data representing, relative to a control, a characteristic for each individual gene as a function of experimental condition, comprising the steps of:

a)converting the gene characteristic to a numeric representation of that characteristic;

b) identifying linear components and nonlinear components of the genes; and

c) performing a linear correlation on the linear components and a nonlinear correlation on the linear components to identify clusters of data having a common frequency characteristic in the time-frequency domain;

thereby to identify the gene as a member of a functional family.

17. The method of claim 16, where the data representing a characteristic for each individual gene is time-varying, further comprising, after step a), the step of:

a1) performing a time-frequency transform on the numeric representation of the gene characteristic.