The traditional strategies may fall short, given that novel genes have been identified in invertebrates that lack the level of similarity to known sequences required for many of the gene finding programs, and many gene prediction programs perform poorly in the GC-poor sequences [ 5] that may characterize invertebrate genomes [ 16].
Many pseudoknot prediction programs have been evaluated using all the entries in the PseudoBase database (van Batenburg et al., 2000).
Since then, CpG islands have been used as landmarks in many promoter prediction programs [ 9, 29, 27, 13].
Finally, many gene prediction programs such as AUGUSTUS provide filtered lists of exons whose scores are hard to interpret in a probabilistic or statistical way.
Therefore, many promoter prediction programs, such as promoterscan [ 9], TSSG, and TSSW [ 10], have been developed based on the density of TFBS.
Many gene prediction programs allow the identification of genes within a new genome [ 1] and the General Feature Format GFFF, proposed by Durbin and Haussler, http://www.sanger.ac.uk/software/gff/) is often chosen for storing the resulting data [ 2].
Motivation: Promoter prediction is an important task in genome annotation projects, and during the past years many new promoter prediction programs (PPPs) have emerged.
Among many miRNA target prediction programs, TargetScan had the best performance based on two large scale miRNA induced protein synthesis studies [ 63, 64].
In addition to these programs there are many other gene prediction programs that are not necessarily geared for EST data and are usually applied to longer nucleic acid sequences or complete genomes (for a review, see [ 18, 19]).
