Gene Structure Prediction and Alternative Splicing Analysis Using Genomically Aligned ESTs

We have used the standard measures of gene prediction performance (Burset and Guigo 1996; Reese et al. 2000) to evaluateTAP. Sn, sensitivity. Sp, specificity. ME, missed exons. WE, wrong exons. The test set contains 1007 multi-exon genes. At the exon level, the values of Sn and ME were calculated individually, summed up, and averaged over 1007 sequences. The values of Sp and WE were averaged over the number of genes for which a prediction was made (Sn >0). A predicted exon is correct if it has an exact match with a known exon. At the gene level, Sn is the proportion of known gene structures that are correctly predicted. Sp is the proportion of predicted gene structures that are correct. A correct gene structure prediction must have perfect Sn and Sp at the exon level.

↵The evaluation used the inferred gene boundaries. Thus, to be considered correct, a predicted exon must be partitioned into the correct genic region and matched to a known exon.