Integrative Approach to Generate High Profile Gene Prediction, the Vitis Genome Case

Thanks to the affordable prices, relative ease and speed of execution, next generation sequencing initiated a rush on genomes being sequenced in the past recent years. The next step after assembly is the gene prediction. Often an underestimated step, that if done right, opens doors for all downstream analyses.

Here we will elaborate on how we proceeded in generating a high quality annotation for Vitis vinifera, by a priori integrating knowledge from other plant genomes in the pipeline, informing the prediction software EuGene about features along the genomic sequence such as protein homology or proof of transcription from ESTs or RNA-seq. This extrinsic information is then used to score transitions along tracks in a direct acyclic graph (DAG) already containing the ab initio elements needed for prediction. In this DAG, a path is searched with the best overall score along the genomic sequence, resulting in the best gene models.

But even with all the efforts done to obtain the best possible prediction, prediction remains only a prediction. Human curation is needed to purify these predictions from flaws and difficult cases where the gene-caller failed.

To allow curation and community annotation we developed ORCAE. A web based system allowing text editing, to add functional description and knowledge, on particular genes of interest, but also a graphical interface to edit genes directly on the sequence, aided by RNA-seq, EST and protein alignments.