Sequence determinants in human polyadenylation site selection

BMC Genomics

Table 2 Negative predictions and accuracy of the ERPIN and POLYADQ program, evaluated for different control sequences not containing polyadenylation sites: coding sequences (CDS), introns, and two types of randomized UTR sequences: simple shuffling or first order Markov simulation.

Negative set	AWUAAA per 100 kb	Program	TN	FP	FP per 100 kb	Specificity SP	Accuracy CC
CDS	31.2	Erpin	880	102	3.7	84.33 %	0.483
		Polyadq	862	120	3.8	82.01 %	0.459
Introns	156.4	Erpin	741	241	38.9	69.49 %	0.320
		Polyadq	718	264	42.0	67.45 %	0.293
UTR shuffled	109.6	Erpin	888	94	11.0	85.38 %	0.494
		Polyadq	826	156	17.4	77.81 %	0.415
UTR Markov 1 ^st order	94.49	Erpin	772	210	21.9	72.33 %	0.354
		Polyadq	733	249	23.9	68.72 %	0.309

See Methods for information on database construction. Each row shows the number of potential A(A/U)UAAA signals per 100 kb in the dataset, True Negatives (TN), False Positives (FP), False Positives per 100 kb, Specificity (SP) and Accuracy (CC). Calculation of CC uses TP and TN from Table 1.

ISSN: 1471-2164