Table 4.

Whole genome promoter prediction in S. cerevisiae using “PromPredict”.

	Number of sequences	Transcript median length	TP	FP	TP promoters	Precision	Recall	F-score
ORF	4912	1548	5317	9072	3934	37.0	80.1	50.6
Cuts	501	428	540	242	404	69.1	80.6	74.4
Suts	729	964	727	704	552	50.8	75.7	60.8
Other	300	1272	296	447	223	39.8	74.3	51.9
All	6442	1436	6880	10465	5113	39.7	79.4	52.9
TATA and TATA-less promoters
TATA	842	1384	978	1298	701	43.0	83.3	56.7
TATA-less	4070	1544	4206	7529	3139	35.8	77.1	48.9

Promoter prediction for 16 chromosomes for both forward and reverse strands has been carried out. −500 to +100 region relative to transcript start was chosen as a true positive region. The performance of PromPredict has been evaluated using the parameters precision, recall, and F-score. Precision is the ratio of number of true positives to the sum of true and false positive predictions, while recall is the ratio of the numbers of promoters with an identified true positive gene to the total number of promoters. TATA-containing and TATA-less gene promoters are defined based on the criterion of presence of TATA-box in −150 to −1 region relative to TSS²². Recall values for prediction of different transcripts belonging to ORF, non-protein coding CUTs, SUTs and other RNA classes (tRNA, rRNA, and SnoRNA) suggest that PromPredict is a good predictor for yeast promoter sequences. The algorithm performs better for TATA-containing gene promoters as compared to TATA-less promoters.