Gene Prediction in Genomic DNA

In about half the cases, the computationally predicted genes were identical to the EST aligmnents 25% of the genes were predicted with less accuracy, and the remaining 25% were predicted poorly. In some cases, 5 sequences from ESTs showed that the gene predictions were either too long or too short. 

Comparisons of the EST sequences to the genomic sequence confirm that the error rate of the worm genome sequence is less than 1 mistake per 10,000 nucleotides. Instances where many ESTs share identical sequence may indicate errors in genomic sequence. Alternatively, these differences could be sequence polymorphisms. 

About 30% of the ESTs exist in alternatively-spliced forms. Many of these alternative splices are not aimotated on the genomic sequence. 

Computational methods may predict separate genes, whereas EST analysis shows that these segments are actually exons of a single gene. Conversely, the computational method may predict exons in cases that should be separate genes. 

---