Skip to main content

Table 2 Comparison of the performance of LocateP with other SCL prediction tools. The entry in each cell indicates the recall of the method with respect to the data in the test-set (TS). * indicates that the test data were extracted from experimental studies. N/A indicates that a certain tool was not applied to the test sets because that set could not be treated appropriately by the tool. The size of the test sets (TS) is indicated in brackets and the relevant literature is mentioned in the Table legend.

From: LocateP: Genome-scale subcellular-location predictor for bacterial proteins

Comparison of LocateP and other SCL prediction tools

Methods

TS 1 (171)

TS 2 (1077)

TS 3 (236)

TS 4 (36)

TS 5 (78)

TS 6 (43)

TS 7 (47)

TS 8 (103)

LocateP

98.8%

99.4%

97.5%

97.2%

91.0%

95.7%

97.9%

98.1%

LipoP

N/A

96.8%

N/A

N/A

89.4%

95.7%

N/A

SignalP 3.0-NN

99.3%

98.3%

97.2%

25.6%

N/A

N/A

N/A

SignalP 3.0-HMM

99.4%

96.6%

97.2%

20.5%

N/A

N/A

N/A

Phobius

98.8%

96.6%

97.2%

42.3%

N/A

N/A

96.1%

Predisi

99.4%

93.2%

94.4%

37.2%

N/A

N/A

N/A

TMHMM

N/A

99.3%

N/A

N/A

N/A

N/A

N/A

97.1%

Psortb v.2.0

N/A

N/A

49.2%

36.1%

10.3% (M)

1.3% (E)

18.6% (M)

4.7% (E)

10.6% (M)

4.3% (E)

N/A

Cello

N/A

N/A

82.6%

80.6%

75.6% (M)

8.0% (E)

61.7% (M)

16.3% (E)

68.1% (M)

27.7% (E)

N/A

Comparison of LocateP, Tat-find v1.2 and TatP in the prediction of Tat-secreted proteins

Methods

TS 3+ TS 4 (272)

TS 9a (713)

TS 9b (632)

TatP

92.8%

99.6%

96.5%

Tat-find v1.2

94.9%

98.6%

93%

LocateP

93.6%

99.9%

98.4%

Comparison of LocateP, Cello and Psortb v2.0 based on data sets extracted from Swiss-Prot

Methods

TS10a (196)

TS10b(129)

TS10c(108)

TS10d(14)

TS11a (340)

TS11b(60)

TS11c(402)

TS11d(50)

LocateP

98%

97%

80.6%e

84%f

97.4%

96.7%

86.1%

86%g

Psortb v.2.0

93.9%

91.7%

79.6%

50%

89.1%

6.7%

81.1%

80%

CELLOd

97%

99.2%

97.2%

57.1%

94.1%

56.7% (E)

43.3% (M)

87.6%

94%

TBPredh

N/A

N/A

N/A

N/A

94.71%

68.33%

87.81%

50%

  1. The test sets are: TS1 [175], TS2 [98]NGP = Cytoplasmic; TS3 [98]PGP, TS4 [41]* = Secreted; TS5 [41]*a = N-anchored; TS6 [157]*, TS7 [151]c [41]b* = Lipid-anchored; TS8 [175] = Membrane; TS9a [86]TestRR = Cytoplasmic; TS9b [86]TestRR = Membrane; TS 10a [28]Test,Training = Cytoplasmic; TS 10b [28]Test,Training = Membrane; TS 10c [28]Test,Training = Extracellular; TS 10d [28]Test,Training = Cell wall; TS11a [95]Training = Cytoplasmic; TS11b [95]Training = Lipid-anchored; TS11c [95]Training = Membrane; TS11d [95]Training = Secreted.
  2. Abbreviations: TS: test set; M: Membrane; E: Extracellular; Test: test set of this article; Training: training set of this article; NGP: negative training set containing only Gram-positive bacterial proteins; PGP: positive training set containing only Gram-positive bacterial proteins; RR: the proteins contain twin-arginine residues in the initial 35 residues.
  3. a: 30 proteins of this set contained putative SPI-cleavage site and were included in LocateP training process
  4. b: After removing redundancy, 47 proteins were left in this set
  5. c: The set contains both Gram-positive and Gram-negative bacterial proteins
  6. d: Only the predictions with highest score were taken
  7. e: 17 proteins in this test set were either proven to be secreted or they were found to be secreted via minor secretion pathways. LocateP focuses on the prediction of major secretion systems, therefore these proteins were predicted as "intracellular", which meant that no classical signal peptides were found in these proteins.
  8. f: Most of the proteins in this set are associated on the cell wall via non-covalent interactions such as protein-protein interaction.
  9. g: 23 out of 50 proteins in this set were predicted as "N-anchored" proteins by LocateP, indicating that these proteins could be secreted via Sec-pathway but remained attached to the cytoplasmic membrane of the cell.
  10. h: among the support-vector machines involved in TBPred only the best performance with the appropriate protein class was taken.
Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us