GENEID - A COMPUTER SERVER FOR PREDICTION OF GENES IN DNA SEQUENCES

We have developed a system, GenelD, for prediction of coding sequences in vertebrate DNA sequences. The approach relies on an ensemble of simple algorithms, the output of which is hierarchically assembled into likely models of gene structure. When tested on vertebrate genes smaller that 15000 basepairs, the actual coding sequence is predicted with an average correlation coefficient of 0.70. This accuracy is higher than any other method currently available. On average, 69% of the coding region is predicted, which in most cases will be sufficient for a search against protein databases. GenelD is available as an electronic mail server, and the predicted gene is automatically compared to protein databases, and any similarities to known proteins reported.