Table 1.
Prediction Performance With Different Pieces of Evidence
| Number of exons | Correct exons | Approx. exons | bSN | bSP | eSN | eSP | ME | WE | |
|---|---|---|---|---|---|---|---|---|---|
| AN | 386 | ||||||||
| GS | 417 | 272 | 286 | 0.96 | 0.91 | 0.63 | 0.61 | 0.03 | 0.08 |
| AS | 838 | 95 | 137 | 0.87 | 0.66 | 0.17 | 0.10 | 0.04 | 0.47 |
| TS | 375 | 58 | 72 | 0.92 | 0.75 | 0.19 | 0.18 | 0.09 | 0.17 |
| BS | 185 | 32 | 42 | 0.45 | 0.68 | 0.16 | 0.18 | 0.51 | 0.34 |
| GS + AS | 446 | 284 | 311 | 0.98 | 0.92 | 0.71 | 0.66 | 0.02 | 0.12 |
| GS + TS | 413 | 282 | 306 | 0.98 | 0.94 | 0.70 | 0.68 | 0.03 | 0.08 |
| GS + TPS | 417 | 278 | 304 | 0.98 | 0.94 | 0.70 | 0.67 | 0.03 | 0.10 |
| GS + BS | 429 | 278 | 304 | 0.98 | 0.92 | 0.69 | 0.66 | 0.03 | 0.11 |
| GS + TS + BS | 415 | 281 | 306 | 0.98 | 0.95 | 0.70 | 0.68 | 0.03 | 0.08 |
Performance comparison of gene identification with GS(GENSCAN), AS (base score), TS (phase-independent TBLASTX score), BS (BLASTX score), GS + AS (GENSCAN and base score), GS + TS (GENSCAN and phase-independent TBLASTX score), GS + TPS (GENSCAN and phase-dependent TBLASTX score), GS + BS (GENSCAN and BLASTX score), and GS + TS + BS (GENSCAN and phase-independent TBLASTX score and BLASTX score). The performance is evaluated by leave-one-out cross-validation. The model is trained on all but one sequence and then evaluated on the remaining sequence. Performance is finally averaged over all such possible partitions. (AN) Annotation. (Correct exons) Predicted exons whose both boundaries are correctly predicted. (Approx exons) Predicted exons with both boundaries close to the boundaries of annotated exons (<10 nts). (bSN) base sensitivity, (bSP) base specificity, (eSN) exon sensitivity, (eSP) exon specificity, (ME) missing exons, and (WE) wrong exons are estimated following Burset and Guigo (1996).