An Incremental Updating Based Fast Phenotype Structure Learning Algorithm

Unsupervised phenotype structure learning is important in microarray data analysis. The goal is to (1) find groups of samples corresponding to different phenotypes (e.g. disease or normal), and (2) find a subset of genes that can distinguish different groups. Due to the large number of genes and a mass of noise in microarray data, the existing methods are often of some limitations in terms of efficicency and effectiveness. In this paper, we develop an incremental updating based phenotype structure learning algorithm, namely FPLA. With a randomly selected initial state, the algorithm iteratively tries three possible adjustments, i.e. gene addition, gene deletion and sample move, to improve the quality of the current result. Accordingly, four incremental updating based optimization strategies are devised to eliminate the redundancy computations in each iteration. Further, by utilizing a harmonic quality function, it improves the result accuracy by penalizing the “outlier” effect. The experiments conducted on several real microarray datasets show that FPLA outperforms the two representative competing algorithms on both effectiveness and efficiency.

[1]  C. D. Gelatt,et al.  Optimization by Simulated Annealing , 1983, Science.

[2]  Daniel R. Rhodes,et al.  CIT: identification of differentially expressed clusters of genes from microarray data , 2002, Bioinform..

[3]  Guoren Wang,et al.  Finding Novel Diagnostic Gene Patterns Based on Interesting Non-redundant Contrast Sequence Rules , 2011, 2011 IEEE 11th International Conference on Data Mining.

[4]  J. Mesirov,et al.  Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. , 1999, Science.

[5]  Alfonso Valencia,et al.  A hierarchical unsupervised growing neural network for clustering gene expression patterns , 2001, Bioinform..

[6]  Philip S. Yu,et al.  /spl delta/-clusters: capturing subspace correlation in a large data set , 2002, Proceedings 18th International Conference on Data Engineering.

[7]  Jian Pei,et al.  Mining phenotypes and informative genes from gene expression data , 2003, KDD '03.

[8]  Elias Zintzaras,et al.  Forest classification trees and forest support vector machines algorithms: Demonstration using microarray data , 2010, Comput. Biol. Medicine.

[9]  Todd,et al.  Diffuse large B-cell lymphoma outcome prediction by gene-expression profiling and supervised machine learning , 2002, Nature Medicine.

[10]  E. Dougherty,et al.  Gene-expression profiles in hereditary breast cancer. , 2001, The New England journal of medicine.

[11]  Jorng-Tzong Horng,et al.  An expert system to classify microarray gene expression data using gene selection by decision tree , 2009, Expert Syst. Appl..

[12]  J. Thomas,et al.  An efficient and robust statistical modeling approach to discover differentially expressed genes using genomic expression profiles. , 2001, Genome research.

[13]  Zhiwen Yu,et al.  Graph-based consensus clustering for class discovery from gene expression data , 2007, Bioinform..

[14]  William M. Rand,et al.  Objective Criteria for the Evaluation of Clustering Methods , 1971 .

[15]  Ge Yu,et al.  Maximal Subspace Coregulated Gene Clustering , 2008, IEEE Transactions on Knowledge and Data Engineering.

[16]  R. Tibshirani,et al.  Supervised harvesting of expression trees , 2001, Genome Biology.

[17]  P. Törönen,et al.  Analysis of gene expression data using self‐organizing maps , 1999, FEBS letters.

[18]  Vipin Kumar,et al.  Subspace Differential Coexpression Analysis: Problem Definition and a General Approach , 2010, Pacific Symposium on Biocomputing.