Clustering Gene Expression Profiles with Memetic Algorithms
- Cite this paper as:
- Merz P., Zell A. (2002) Clustering Gene Expression Profiles with Memetic Algorithms. In: Guervós J.J.M., Adamidis P., Beyer HG., Schwefel HP., Fernández-Villacañas JL. (eds) Parallel Problem Solving from Nature — PPSN VII. PPSN 2002. Lecture Notes in Computer Science, vol 2439. Springer, Berlin, Heidelberg
Microarrays have become a key technology in experimental molecular biology. They allow a monitoring of gene expression for more than ten thousand genes in parallel producing huge amounts of data. In the exploration of transcriptional regulatory networks, an important task is to cluster gene expression data for identifying groups of genes with similar patterns.
In this paper, memetic algorithms (MAs) — genetic algorithms incorporating local search — are proposed for minimum sum-of-squares clustering. Two new mutation and recombination operators are studied within the memetic framework for clustering gene expression data. The memetic algorithms using a sophisticated recombination operator are shown to converge very quickly to (near-)optimum solutions. Furthermore, the MAs are shown to be superior to multi-start k-means clustering algorithms in both computation time and solution quality.
Unable to display preview. Download preview PDF.