Table 1

Performance of ExportPred variants

Model number

1

2

3

4

PEXEL WMM

Default

Default

Hiller

Hiller

Signal sequence model

Default

SignalP

Default

SignalP

Negative set

Positive set: training sequences

PfNegative

0.98

0.90

0.97

0.88

Simulated

0.99

0.95

0.99

0.95

Simulated (PfSS)

0.96

0.80

0.92

0.61

Simulated (SpSS)

0.97

0.50

0.95

0.16

Simulated (EPSS)

0.95

0.88

0.91

0.79

Negative set

Positive set: Rifins + Stevors

PfNegative

0.96

0.97

0.95

0.94

Simulated

0.98

0.99

0.97

0.99

Simulated (PfSS)

0.95

0.93

0.91

0.77

Simulated (SpSS)

0.95

0.61

0.93

0.21

Simulated (EPSS)

0.91

0.96

0.89

0.91


Performance of ExportPred as measured by area under the respective ROC curve for combinations of model variant, and positive and negative dataset. For each pair of positive and negative sets, the best performing model is highlighted in bold. The four model variants are constructed by substituting ExportPred PEXEL weight model matrix (WMM) with the one published in [6,7] and/or by substituting the ExportPred signal sequence states with the HMM used in SignalP.

Sargeant et al. Genome Biology 2006 7:R12   doi:10.1186/gb-2006-7-2-r12

Open Data