Recent trend in gene prediction is to combine similarity information with ab initio predictions (Ashurst and Collins, 2003). Presence of similar sequences in databses increases the probability of query sequences being correctly predicted. Combination of ab initio gene predictions with similarity information has previously resulted in improvement of one aspect of prediction accuracy. The increase in coverage or sensitivity is more or less offset by the reduction in specificity. EGPred implements a new strategy where the specificity is maintained while the sensitivity is increased on combining ab initio predictions with BLASTX exons that are filtered using criteria such as length of HSP alignment, percent identity, score, expectation value, and similarity to intron sequences. This strategy increases the sensitivity from 0.95 to 0.97 for Genscan, from 0.93 to 0.96 for HMMgene, from 0.94 to 0.96 for EUI method, from 0.93 to 0.96 for EUI-Frame method, and from 0.91 to 0.95 for GI method on HMR195 dataset using default values. The specificity at the nucleotide level remain unchanged due to the combination procedure, thereby fulfilling a major requirement of researchers for a more accurate predictions from combination methods.

A comprehensive and detailed evaluation of the EGPred server is available elsewhere.

Send Comments and Suggestions to: raghava@iiitd.ac.in
If you use EGPred, please site:
Issac, B and Raghava, G P S (2004), EGPred: Prediction of Eukaryotic Genes Using ab initio Methods After Combining With Sequence Similarity Approaches, Genome Res 14, 1756-1766.

---