GLEAN is an unsupervised learning system to integrate disparate sources of gene structure evidence (gene model predictions, EST/protein genomic sequence alignments, SAGE/peptide tags, etc) to produce a consensus gene prediction, without prior training.