Abstract
More than half of the Arabidopsis genes are functionally unknown or annotated just based on sequence similarities. It's important to support functional genomics with massive "-omics" information like high-throughput sequencer or genome tiling array results and bioinformatics analyses. In this study, we constructed a new method named "DSDE". We simultaneously constructed gene structures from gene expression observations (Dynamic Structure Construction, with a program named ARTADE2) and predicted gene functions with the Dynamic Gene Expression analysis. We applied this method to the Arabidopsis transcriptomes and got the following results. (1)We got about 90% success rates against well-annotated genes for re-prediction of original annotation terms. (2) We put some function predictions on 98% of unknown or poorly annotated genes. (3) We showed about 1500 novel gene candidates with their function predictions. Here, we published a database named "ARTADE2DB" serving these results.
ARTADE2DB: http://scinets.org/db/artade2