Evolutionary Clustering Algorithm for Gene Selection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current clustering algorithms for gene expression data face limitations such as sensitivity to outliers, initial cluster number specification, and biased distribution, which affect the identification of candidate genes for classifying diseased and normal states.

Innovation Solution

The Evolutionary Clustering Algorithm (ECA) combines clustering and genetic algorithms, using an interdependence redundancy measure for gene data clustering and selection, which addresses these limitations by employing a hybrid approach to identify co-regulated and co-expressed genes effectively.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If traditional clustering algorithms (K-means, SOM) are used with conventional distance measures, then the clustering process is simple and fast, but the measures are sensitive to outliers and require initial specification of cluster number

Engineering Contradiction:

Improveclustering process simplicityVSAvoidclustering accuracy

Core Design Contradiction:

Ease of operationVSReliability

Solution Approach 1:

The patent changes the distance measurement parameter from conventional Euclidean distance or Pearson correlation to an information-theoretic measure (interdependence redundancy measure). This parameter change makes the clustering more robust to outliers and eliminates the need for initial cluster number specification, while maintaining computational feasibility through the hybrid evolutionary approach

Inventive Principle:

Principle #35Parameter changes

2Productivity

If the number of clusters is specified initially in traditional clustering algorithms, then the algorithm execution is straightforward, but it limits the ability to discover optimal cluster configurations

Engineering Contradiction:

Improvealgorithm execution efficiencyVSAvoidcluster configuration flexibility

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent makes the cluster number dynamic rather than static. The hybrid evolutionary algorithm dynamically determines the optimal number of clusters during the clustering process by evaluating multiple possible configurations, allowing the system to adapt to the data structure without requiring predetermined cluster counts

Inventive Principle:

Principle #15Dynamics

3Device complexity

If conventional distance measures are used for gene clustering, then the computation is straightforward, but the measures fail to capture interdependence between genes

Engineering Contradiction:

Improvecomputation complexityVSAvoidgene interdependence information

Core Design Contradiction:

Device complexityVSLoss of information

Solution Approach 1:

The patent substitutes the mechanical distance-based clustering approach with an information-theoretic approach. Instead of using geometric distance measures that only capture linear relationships, the invention employs interdependence redundancy measure based on information theory to capture complex non-linear interdependencies between genes, thereby reducing information loss about gene relationships

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS8712935B2Evolutionary clustering algorithm

Publication Date: 2014.04.29 KONINKLIJKE PHILIPS NV

AI summary

The invention relates to selecting a set of candidate genes from a pool of genes. The method comprising receiving a set of gene data; arranging the set of gene data into a set of clusters with similar profiles by use of a clustering algorithm; and inputting the set of clusters into a genetic algorithm to select a set of candidate genes from the set of clusters. The method thus relates to hybrid between selection by clustering computation and selection by evolutionary computation. This hybrid is also referred to as an evolutionary clustering algorithm (ECA).