Inferring Meta-covariates in Classification

Authors:
Keith Harris;Lisa Mcmillan;Mark Girolami
Affiliations:
Inference Group, Department of Computing Science, University of Glasgow, UK;Inference Group, Department of Computing Science, University of Glasgow, UK;Inference Group, Department of Computing Science, University of Glasgow, UK
Venue:
PRIB '09 Proceedings of the 4th IAPR International Conference on Pattern Recognition in Bioinformatics
Year:
2009

Citing 4
Cited 0

Improving classification of microarray data using prototype-based feature selection

ACM SIGKDD Explorations Newsletter
Gene selection using a two-level hierarchical Bayesian model

Bioinformatics
Pattern Recognition and Machine Learning (Information Science and Statistics)

Pattern Recognition and Machine Learning (Information Science and Statistics)
Bayesian Regularization for Normal Mixture Estimation and Model-Based Clustering

Journal of Classification

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper develops an alternative method for gene selection that combines model based clustering and binary classification. By averaging the covariates within the clusters obtained from model based clustering, we define "meta-covariates" and use them to build a probit regression model, thereby selecting clusters of similarly behaving genes, aiding interpretation. This simultaneous learning task is accomplished by an EM algorithm that optimises a single likelihood function which rewards good performance at both classification and clustering. We explore the performance of our methodology on a well known leukaemia dataset and use the Gene Ontology to interpret our results.