Methods for classifying samples and ascertaining previously unknown classes

• US 6,647,341 B1
• Filed: 04/06/2000
• Issued: 11/11/2003
• Est. Priority Date: 04/09/1999
• Status: Expired due to Fees

• Current Assignee: Dana-Farber Institute, Inc., Whitehead Institute for Biomedical Research
• Original Assignee: Dana-Farber Cancer Institute, Whitehead Institute for Biomedical Research
• Inventors: Golub, Todd R., Slonim, Donna, Mesirov, Jill, Lander, Eric S., Tamayo, Pablo
• Primary Examiner(s): Woodward, Michael P.
• Assistant Examiner(s): Clow, Lori A.
• Application Number:
  US09/544,627
• Time in Patent Office: 1,314 Days
• Field of Search: 702/19, 702/20, 435/6
• US Class Current: 702/19
• 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

Abstract

Methods and apparatus for classifying or predicting the classes for samples based on gene expression are described. Also described are methods and apparatus for ascertaining or discovering new, previously unknown classes based on gene expression.

Claims

1. A method of identifying a set of informative genes whose expression correlates with a class distinction between samples, comprising the steps of:

a. sorting genes using a neighborhood analysis, wherein the genes are sorted by degree to which their expression in said samples correlate with a class distinction; and

b. determining whether said correlation is stronger than expected by chance;

wherein genes whose expression correlate with a class distinction more strongly than expected by chance are informative genes, thereby identifying a set of informative genes.

2. The method of claim 1, wherein the class distinction is determined between known classes.

3. The method of claim 2, wherein the class distinction is a disease class distinction.

4. The method of claim 3, wherein the disease class distinction is a cancer class distinction.

5. The method of claim 4, wherein the cancer class distinction is selected from the group consisting of a leukemia class distinction, a brain tumor class distinction and a lymphoma class distinction.

6. The method of claim 1, wherein said neighborhood analysis comprises the steps of:

a. defining an idealized expression pattern corresponding to a gene, wherein said idealized expression pattern is expression of said gene that is uniformly high in a first class and uniformly low in a second class; and

b. determining whether there is a high density of genes having an expression pattern similar to said idealized expression pattern, as compared to an equivalent random expression pattern, wherein the high density of genes are genes having a high statistical significance in a permutation test.

7. The method of claim 6, further including a signal to noise routine performed according to:

8. A method of identifying a set of informative genes whose expression correlates with a class distinction between samples, comprising the steps of:

a. sorting genes using a neighborhood analysis, wherein the genes are sorted by degree to which their expression in said samples correlate with a disease class distinction; and

b. determining whether said correlation is stronger than expected by chance;

wherein genes whose expression correlate with a disease class distinction more strongly than expected by chance are informative genes, thereby identifying a set of informative genes.

9. A method of identifying a set of informative genes whose expression correlates with a class distinction between samples, comprising the steps of:

a. sorting genes using a neighborhood analysis, wherein the genes are sorted by degree to which their expression in said samples correlate with a cancer disease class distinction; and

b. determining whether said correlation is stronger than expected by chance;

wherein genes whose expression correlate with a cancer disease class distinction more strongly than expected by chance are informative genes, thereby identifying a set of informative genes.

10. A method of identifying a set of informative genes whose expression correlates with a class distinction between samples, comprising the steps of:

a. defining an idealized expression pattern corresponding to a gene, wherein said idealized expression pattern is expression of said gene that is uniformly high in a first class and uniformly low in a second class; and

b. determining whether there is a high density of genes having an expression pattern similar to said idealized expression pattern, as compared to an equivalent random expression pattern, wherein the high density of genes are genes having a high statistical significance in a permutation test;

wherein genes whose expression correlate with a disease class distinction more strongly than expected by chance are informative genes, thereby identifying a set of informative genes.

11. The method of claim 10, wherein Step a. is performed by a signal to noise routine according to:

12. A method of identifyng a set of informative genes whose expression correlates with a class distinction between samples, comprising the steps of:

a. sorting genes using a neighborhood analysis, wherein the genes are sorted by degree to which their expression in said samples correlate with a cancer disease class distinction, said neighborhood analysis includes performing a signal to noise routine according to;