Discovering Knowledge from Local Patterns in SAGE Data

Discovering Knowledge from Local Patterns in SAGE Data

Bruno Crémilleux (Université de Caen, France), Arnaud Soulet (Université François Rabelais de Tours, France), Jiri Kléma (Czech Technical University in Prague, Czech Republic), Céline Hébert (Université de Caen, France) and Olivier Gandrillon (Université de Lyon, France)
DOI: 10.4018/978-1-60566-218-3.ch012
OnDemand PDF Download:


The discovery of biologically interpretable knowledge from gene expression data is a crucial issue. Current gene data analysis is often based on global approaches such as clustering. An alternative way is to utilize local pattern mining techniques for global modeling and knowledge discovery. Nevertheless, moving from local patterns to models and knowledge is still a challenge due to the overwhelming number of local patterns and their summarization remains an open issue. This chapter is an attempt to fulfill this need: thanks to recent progress in constraint-based paradigm, it proposes three data mining methods to deal with the use of local patterns by highlighting the most promising ones or summarizing them. Ideas at the core of these processes are removing redundancy, integrating background knowledge, and recursive mining. This approach is effective and useful in large and real-world data: from the case study of the SAGE gene expression data, we demonstrate that it allows generating new biological hypotheses with clinical application.
Chapter Preview


In many domains, such as gene expression data, the critical need is not to generate data, but to derive knowledge from huge and heterogeneous datasets produced at high throughput. It means that there is a great need for automated tools helping their analysis. There are various methods, including global techniques such as hierarchical clustering, K-means, or co-clustering (Madeira & Oliveira, 2004) and approaches based on local patterns (Blachon et al., 2007). In the context of genomic data, a local pattern is typically a set of genes displaying specific expression properties in a set of biological situations. A great interest of local patterns is to capture subtle relationships in the data which are not detected by global methods and leading to the discovery of precious nuggets of knowledge (Morik et al., 2005). But, the toughness of extraction of various local patterns is a substantial limitation of their use (Ng et al.,1998; Bayardo, 2005). As the search space of the local patterns exponentially grows according to the number of attributes (Mannila &Toivonen, 1997), this task is even more difficult in large datasets (i.e., datasets where objects having a large number of columns).This is typically the case in gene expression data: few biological situations (i.e., objects) are described by ten of thousands of gene expressions values (i.e., attributes) (Becquet et al. 2002). In such situations, naive methods or usual level-wise techniques are unfeasible (Pan et al.,2003; Rioult et al., 2003). Nevertheless, especially in the context of transactional data, the recent progress in constraint-based pattern mining (see for instance (Bonchi & Lucchese, 2006; De Raedt et al., 2002) enable to extract various kind of patterns even in large datasets (Soulet et al., 2007). But, this approach has still a limitation: it tends to produce an overwhelming number of local patterns. Pattern flooding follows data flooding: the output is often too large for an individual and global analysis performed by the end-user. This is especially true in noisy data,such as genomic data where the most significant patterns are lost among too many trivial, noisy and redundant information. Naive techniques such as tuning parameters of methods (e.g., increasing the frequency threshold) limit the output but only lead to produce trivial and useless information.

This paper tackles this challenge. Relying on recent progress in constraint-based paradigm, it presents three data mining methods to deal with the use of local patterns by highlighting the most promising ones or summarizing them. The practical usefulness of these methods are supported by the case study of the SAGE gene expression data (introduced in the next section). First, we provide a method to mine the set of the simplest characterization rules while having a controlled number of exceptions. Thanks to their property of minimal premise, this method limits the redundancy between rules. Second, we describe how to integrate in the mining process background knowledge available in literature databases and biological ontologies to focus on the most promising patterns only. Third, we propose a recursive pattern mining approach to summarize the contrasts of a dataset: only few patterns conveying a trade-off between significance and representativity are produced. All of these methods can be applied even on large data sets. The first method comes within the general framework of removing redundancy and providing lossless representations whereas the two others propose summarizations (all the information cannot be regenerated but the most meaningful features are produced). We think that these two general approaches are complementary. Finally, we sum up the main lessons coming from mining and using local patterns on SAGE data, both from the data mining and the biological points of view. It demonstrates the practical usefulness of these approaches enabling to infer new relevant biological hypotheses.

This paper abstracts our practice of local patterns discovery from SAGE data. We avoid technical details (references are given for in-depth information), but we emphasize the main principles and results and we provide a cross-fertilization of our “in silico” approaches for discovering knowledge in gene expression data from local patterns.

Complete Chapter List

Search this Book:
Editorial Advisory Board
Table of Contents
Riccardo Bellazzi
Petr Berka, Jan Rauch, Djamel Abdelkader Zighed
Petr Berka, Jan Rauch, Djamel Abdelkader Zighed
Chapter 1
Jana Zvárová, Arnošt Veselý
This chapter introduces the basic concepts of medical informatics: data, information, and knowledge. Data are classified into various types and... Sample PDF
Data, Information and Knowledge
Chapter 2
Michel Simonet, Radja Messai, Gayo Diallo
Health data and knowledge had been structured through medical classifications and taxonomies long before ontologies had acquired their pivot status... Sample PDF
Ontologies in the Health Field
Chapter 3
Alberto Freitas, Pavel Brazdil, Altamiro Costa-Pereira
This chapter introduces cost-sensitive learning and its importance in medicine. Health managers and clinicians often need models that try to... Sample PDF
Cost-Sensitive Learning in Medicine
Chapter 4
Arnošt Veselý
This chapter deals with applications of artificial neural networks in classification and regression problems. Based on theoretical analysis it... Sample PDF
Classification and Prediction with Neural Networks
Chapter 5
Patrik Eklund, Lena Kallin Westin
Classification networks, consisting of preprocessing layers combined with well-known classification networks, are well suited for medical data... Sample PDF
Preprocessing Perceptrons and Multivariate Decision Limits
Chapter 6
Xiu Ying Wang, Dagan Feng
The rapid advance and innovation in medical imaging techniques offer significant improvement in healthcare services, as well as provide new... Sample PDF
Image Registration for Biomedical Information Integration
Chapter 7
ECG Processing  (pages 137-160)
Lenka Lhotská, Václav Chudácek, Michal Huptych
This chapter describes methods for preprocessing, analysis, feature extraction, visualization, and classification of electrocardiogram (ECG)... Sample PDF
ECG Processing
Chapter 8
EEG Data Mining Using PCA  (pages 161-180)
Lenka Lhotská, Vladimír Krajca, Jitka Mohylová, Svojmil Petránek, Václav Gerla
This chapter deals with the application of principal components analysis (PCA) to the field of data mining in electroencephalogram (EEG) processing.... Sample PDF
EEG Data Mining Using PCA
Chapter 9
Darryl N. Davis, Thuy T.T. Nguyen
Risk prediction models are of great interest to clinicians. They offer an explicit and repeatable means to aide the selection, from a general... Sample PDF
Generating and Verifying Risk Prediction Models using Data Mining
Chapter 10
Vangelis Karkaletsis, Konstantinos Stamatakis, Karampiperis, Karampiperis, Pythagoras Karampiperis, Pythagoras Karampiperis
The World Wide Web is an important channel of information exchange in many domains, including the medical one. The ever increasing amount of freely... Sample PDF
Management of Medical Website Quality Labels via Web Mining
Chapter 11
Rainer Schmidt
In medicine, a lot of exceptions usually occur. In medical practice and in knowledge-based systems, it is necessary to consider them and to deal... Sample PDF
Two Case-Based Systems for Explaining Exceptions in Medicine
Chapter 12
Bruno Crémilleux, Arnaud Soulet, Jiri Kléma, Céline Hébert, Olivier Gandrillon
The discovery of biologically interpretable knowledge from gene expression data is a crucial issue. Current gene data analysis is often based on... Sample PDF
Discovering Knowledge from Local Patterns in SAGE Data
Chapter 13
Jirí Kléma, Filip Železný, Igor Trajkovski, Filip Karel, Bruno Crémilleux
This chapter points out the role of genomic background knowledge in gene expression data mining. The authors demonstrate its application in several... Sample PDF
Gene Expression Mining Guided by Background Knowledge
Chapter 14
Pamela L. Thompson, Xin Zhang, Wenxin Jiang, Zbigniew W. Ras, Pawel Jastreboff
This chapter describes the process used to mine a database containing data, related to patient visits during Tinnitus Retraining Therapy. The... Sample PDF
Mining Tinnitus Database for Knowledge
Chapter 15
Dinora A. Morales, Endika Bengoetxea, Pedro Larrañaga
Infertility is currently considered an important social problem that has been subject to special interest by medical doctors and biologists. Due to... Sample PDF
Gaussian-Stacking Multiclassifiers for Human Embryo Selection
Chapter 16
Mining Tuberculosis Data  (pages 332-349)
Marisa A. Sánchez, Sonia Uremovich, Pablo Acrogliano
This chapter reviews the current policies of tuberculosis control programs for the diagnosis of tuberculosis. The international standard for... Sample PDF
Mining Tuberculosis Data
Chapter 17
Mila Kwiatkowska, M. Stella Atkins, Les Matthews, Najib T. Ayas, C. Frank Ryan
This chapter describes how to integrate medical knowledge with purely inductive (data-driven) methods for the creation of clinical prediction rules.... Sample PDF
Knowledge-Based Induction of Clinical Prediction Rules
Chapter 18
Petr Berka, Jan Rauch, Marie Tomecková
The aim of this chapter is to describe goals, current results, and further plans of long-time activity concerning application of data mining and... Sample PDF
Data Mining in Atherosclerosis Risk Factor Data
About the Contributors