Mining concepts from large SAGE gene expression matrices

Abstract

Abstract. One of the crucial needs in post-genomic research is to analyze expression matrices (e.g., SAGE and microarray data) to identify a priori interesting sets of genes, e.g., sets of genes that are frequently co-regulated. Such matrices provide expression values for given biological situations (the lines) and given genes (columns). The inductive database framework enables to support knowledge discovery processes by means of sequences of queries that concerns both data processing and pattern querying (extraction, post-processing). We provide a simple formalization of a relevant pattern domain (language of patterns, evaluation functions and primitive constraints) that has been proved useful for specifying various analysis tasks. Recent algorithmic results w.r.t. the efficient evaluation (constraint-based mining) of the so-called inductive queries are emphasized and illustrated on a 90 × 12 636 human SAGE expression matrix. 1