BioLattice: a framework for the biological interpretation of microarray gene expression data using concept lattice analysis

Abstract

MOTIVATION: A challenge in microarray data analysis is to interpret observed changes in terms of biological properties and relationships. One powerful approach is to make associations of gene expression clusters with biomedical ontologies and/or biological pathways. However, this approach evaluates only one cluster at a time, returning long unordered lists of annotations for clusters without considering the overall context of the experiment under investigation. RESULTS: BioLattice is a mathematical framework based on concept lattice analysis for the biological interpretation of gene expression data. By considering gene expression clusters as objects and associated annotations as attributes and by using set inclusion relationships BioLattice orders them to create a lattice of concepts, providing an 'executive' summary of the experimental context. External knowledge resources such as Gene Ontology trees and pathway graphs can be added incrementally. We propose two quantitative structural analysis methods, 'prominent sub-lattice' and 'core-periphery' analyses, enabling systematic comparison of experimental concepts and contexts. BioLattice is implemented as a web-based utility using Scalable Vector Graphics for interactive visualization. We applied it to real microarray datasets with improved biological interpretations of the experimental contexts.