Method and system for determining an association of biological feature with medical condition

Patent metadata

Field	Value
Publication number	US-2017329913-A1
Application number	US-201615150732-A
Country	US
Kind code	A1
Filing date	May 10, 2016
Priority date	May 10, 2016
Publication date	Nov 16, 2017
Grant date	—

Abstract

A method and a system for determining an association of at least one biological feature with a medical condition utilizes the novel L 1/2 penalized network-constraint regression model to achieve an improved biological analysis, in particular by solving high-dimensional problems. The method and the system of the present invention attain high accuracy and preciseness.

First claim

1 . A method of determining an association of at least one biological feature with a medical condition, comprising the steps of:
obtaining a dataset comprising biological data related to a plurality of samples each having a plurality of biological features; applying at least some of the biological data to a regression model to determine and/or optimize parameters in the regression model thereby solving the regression model; and processing the biological data using the solved regression model with a biological model to determine one or more biological features that are associated with the medical condition.

2 . The method in accordance with claim 1 , wherein the regression model includes L 1/2 -regularized regression model.

3 . The method in accordance with claim 1 , wherein the biological model includes biological network information associated with the medical condition and the one or more biological features.

4 . The method in accordance with claim 3 , wherein the biological network information is arranged to represent a biological process, a molecular interaction and/or a reaction network associated with the biological features.

5 . The method in accordance with claim 3 , wherein the step of processing the biological data using the solved regression model with a biological model comprises the step of combining the biological data with the biological network information to determine one or more biological features that are associated with the medical condition.

6 . The method in accordance with claim 3 , wherein the biological network information is graph Laplacian regularized.

7 . The method in accordance with claim 6 , wherein the step of regularizing the biological network information includes an iterative transformation for obtaining at least one estimation representing the correlation between the one or more biological features and the medical condition.

8 . The method in accordance with claim 7 , wherein the iterative transformation includes an univariate half thresholding operation of a coordinate descent optimization of the regularized biological network information for obtaining the model.

9 . The method in accordance with claim 8 , wherein a thresholding representation of

54
3

4



(
λ
)

2
3

is used in the univariate half thresholding operation, wherein λ denotes a regularization parameter.

10 . The method in accordance with claim 8 , wherein the regression model includes L 1/2 penalized network-constrained regression model.

11 . The method in accordance with claim 1 , wherein the at least one biological feature includes at least one of presence of a gene, gene expression, presence of a gene product or amount of a gene product, and the medical condition is cancer.

12 . The method in accordance with claim 11 , wherein the one or more biological features associated with the medical condition includes one or more biomarker and/or indicator arranged to represent an indication of the medical condition.

13 . A system for determining an association of at least one biological feature with a medical condition, comprising a processing module arranged to:
apply at least some of the biological data in a dataset comprising biological data related to a plurality of samples each having a plurality of biological features to a regression model so as to determine and/or optimize parameters in the regression model thereby solving the regression model; and process the biological data using the solved regression model with a biological model to determine one or more biological features that are associated with the medical condition.

14 . The system in accordance with claim 13 , wherein the regression model includes L 1/2 penalized network-constrained regression model.

15 . The system in accordance with claim 13 , wherein the biological model includes biological network information associated with the medical condition and the one or more biological features.

16 . The system in accordance with claim 15 , wherein the one or more processor is further arranged to combine the biological data with the biological network information so as to determine one or more biological features that are associated with the medical condition.

17 . The system in accordance with claim 13 , wherein the biological network information is graph Laplacian regularized.

18 . The system in accordance with claim 17 , wherein the one or more processor is arranged to perform an iterative transformation for obtaining at least one estimation representing the correlation between the one or more biological features and the medical condition.

19 . The system in accordance with claim 18 , wherein the iterative transformation includes an univariate half thresholding operation of a coordinate descent optimization of the regularized biological network information for obtaining the model.

20 . The system in accordance with claim 13 , wherein the at least one biological feature includes at least one of presence of a gene, gene expression, presence of a gene product or amount of a gene product, and the medical condition is cancer, and wherein the one or more biological features associated with the medical condition includes one or more biomarker and/or indicator arranged to represent an indication of the medical condition.

Assignees

• Univ Macau Sci & Tech

Inventors

• Liang Yong
• Huang hai-hui
• Liu xiao-ying

Classifications

• G06N5/022

  Knowledge engineering; Knowledge acquisition · CPC title

• G16H50/20 Primary

  for computer-aided diagnosis, e.g. based on medical expert systems · CPC title

• G06N20/00

  Machine learning · CPC title

• G06N7/023

  Learning or tuning the parameters of a fuzzy system · CPC title

• G06F19/345 Primary

  Physics · mapped topic