Consensus Promoter Identification in the Human Genome Utilizing Expressed Gene Markers and Gene Modeling

Table 2.

Searching Promoters for Genes with Known Complete Coding Region

Programs Training set of 133 promoters Test set of 120 promoters
true prediction (sensitivity) false prediction (FP/AllP) undetected true prediction (sensitivity) false prediction (FP/AllP) undetected
PROSCAN1.7 31 (23%) 21 (40%) 81 26 (22%) 28 (52%) 66
NNPP2.0 63 (47%) 50 (44%) 21 24 (20%) 63 (72%) 33
PromFD1.0 53 (40%) 44 (45%) 48 60 (50%) 63 (51%) 10
promoter2.0 9 (7%) 101 (92%) 23 10 (8%) 92 (90%) 18
TSSG 78 (57%) 13 (14%) 42 56 (47%) 27 (33%) 37
TSSW 58 (44%) 36 (38%) 39 56 (47%) 38 (40%) 26
CONPRO 86 (65%) 9 (9%) 38 75 (63%) 10 (12%) 35
  • The CONPRO can detect promoters for about half of the human genes with mRNA known. For the promoters predicted, about 88% of them are true promoters. Additional information is given in the legend of Table 1.

This Article

  1. Genome Res. 12: 462-469

Preprint Server