Network inferring method

• US 20030097693A1
• Filed: 10/29/2002
• Published: 05/22/2003
• Est. Priority Date: 12/28/1999
• Status: Abandoned Application

• Current Assignee: Japan Science And Technology Agency
• Original Assignee: Japan Science And Technology Agency
• Inventors: Kitano, Hiroaki, Morohashi, Mineo
• Application Number:
  US10/169,433
• Publication Number:
  US 20030097693A1
• Time in Patent Office: Days
• US Class Current: 702/27
• 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 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 5/00 ICT specially adapted for modelling or simulations in systems biology, e.g. gene-regulatory networks, protein interaction networks or metabolic networks

Abstract

The present invention provides a method for predicting a topology stepwise by producing candidate networks that can reproduce a given data profile, evaluating these candidate networks, and predicting a network structure and parameters that can provide the data profile. According to the present invention, a regulatory network for a gene that may induce a certain phenomenon or a metabolic network indicating enzyme and protein reactions can be predicted based on gene expression or protein concentration data.

Claims

1. A network predicting method that can reproduce data from given data produced by mutually related elements when the relationship between the elements can be described as a network, wherein the method comprises the steps of:

a) generating topologies that can reproduce the given data, and for each of the generated topologies, producing a parameter set that has the highest fitness to the given data, thereby obtaining candidate networks;

b) extracting a consensus pathway from the networks produced in step a) that meet a desired criterion;

c) generating topologies containing the consensus pathway extracted in step b) and providing each of the generated topologies with a parameter set that has the highest fitness to the given data, thereby obtaining networks; and

,

d) repeating steps b) and c) using the networks obtained in step c) as candidates until the scale of networks converges to a desired value, thereby obtaining a new network.

2. The network predicting method according to claim 1, wherein the desired criterion is evaluated based on fitness in step b).

3. The network predicting method according to claim 1, wherein the desired criterion is an ability to generate data that are similar to the data generated by a network containing the pathway in step b).

4. The network predicting method according to claim 1, wherein the method comprises generating all topologies that can reproduce the given data in step a).

5. The network predicting method according to claim 1, wherein the method comprises selecting networks that can reproduce the given data in step a), as networks that can reproduce the given data in step a).

6. The network predicting method according to claim 5, wherein the method comprises generating networks that reproduce data having only a small difference from the given data.

7. The network predicting method according to claim 5 or 6, wherein the step of obtaining networks in step a) and/or c) comprises recombining parts of the structures of networks that reproduce data having only small differences from the given data.

8. The network predicting method according to any one of claims 5 to 7, wherein obtaining networks in step a) and/or c) comprises searching structural neighborhoods of an arbitrary network to produce new networks.

9. The network predicting method according to any one of claims 1 to 8, wherein obtaining networks in step a) and/or c) comprises predicting one parameter or a set of several different parameters for the same network.

10. The network predicting method according to any one of claims 1 to 8, wherein obtaining candidate networks in step a) and/or c) comprises predicting parameters using at least one method selected from the group consisting of a genetic algorithm, simulated annealing, and a hill climbing method.

11. The network predicting method according to any one of claims 1 to 10, wherein the data generated by the elements is an expression profile of a gene.

13. A storage medium storing a program that allows a computer system to implement a network predicting method according to any of claims 1 to 11.

12. A network predicting apparatus that can reproduce data from given data produced by mutually related elements when the relationship between the elements can be described as a network, wherein the apparatus comprises the following means:

a) first storage means for storing networks each composed of a topology and a corresponding parameter set;

b) second storage means for storing candidate topologies or elements thereof and a corresponding parameter set;

c) means for generating topologies that can reproduce the given data, and for each of the generated topologies, producing a parameter set that has the highest fitness to the given data, thereby obtaining candidate networks, which are then stored in the first storage means;

d) means for extracting a consensus pathway from candidate networks which meet a desired criterion and are stored in the first storage means, and storing the consensus pathway in the second storage means;

e) means for generating topologies containing the consensus pathway stored in the second storage means by means d), providing each of the generated topologies with a parameter set that has the highest fitness to the given data, thereby obtaining networks, and storing the networks in the first storage means; and

f) means for providing the networks stored in the first storage means by the means e) as candidate networks for the means d) until a network to be predicted is obtained.