Gene finding integrating external evidence

EuG` ene, a gene finder for Arabidopsis thaliana, combines the output of several information sources, including output of other software or user information. To achieve this, a weighted DAG is built in such a way that a shortest feasible path in this graph represents the most likely gene structure of the underlying DNA sequence. The usual simple Bellman linear time shortest path algorithm for DAG has been replaced by a shortest path with constraints algorithm which remains linear both in time and space. EuG` ene eectiveness has been recently assessed on Araset, a recent dataset of Arabidopsis thaliana sequences used to evaluate several existing gene finding software. It appears that, despite its simplicity, EuGene gives results which compare very favorably to existing software.