Abstract : Despite tremendous advances, automatic gene annotations are considered as predictions that require human validation. We have developped Exogean, a software for annotating gene structures in eukaryotic genomic DNA based on human expertise. Exogean explicitely uses the same heuristic rules that human biologists use when annotating genes. Consequently Exogean is conceived as a framework that represents the biological objects (exons, transcripts, etc) and the rules that we use to manipulate them. This framework is based on directed acyclic coloured multigraphs (DACMS), a powerful representation that intuitively models the reasonning followed by human experts.
The fact that the heuristic rules used by human experts may change over time, are difficult to express and are applied to heterogeneous data, raised the need for a generic, flexible and intuitive formal framework for gene annotation. We thus also have developed DACMLang, a language dedicated to gene annotation based on DACMs.