Candidate prioritization for low-abundant differentially expressed proteins in 2D-DIGE datasets

BMC Bioinformatics

Table 1 Prioritization performance

STRING evidence type	Cut-off	TPR ₅	TPR ₁₀	TPR ₁₅	TPR ₂₅
Neighbourhood	0	0.238	0.276	0.305	0.352
Gene fusion	0	0.076	0.095	0.114	0.181
Cooccurrence	0	0.2	0.248	0.257	0.286
Coexpression	0	0.41	0.514	0.571	0.61
Experiments	0	0.295	0.343	0.39	0.457
Database	0	0.229	0.295	0.343	0.39
Textmining	0	0.371	0.438	0.486	0.533
Combined	0	0.438	0.514	0.571	0.6
Combined (gene expression-based filter)	0	0.438	0.514	0.571	0.629
Combined	0.15	0.429	0.514	0.571	0.6
Combined	0.4	0.410	0.476	0.505	0.543
Combined	0.7	0.305	0.4	0.429	0.514

True-positive rates TPR_n estimated using LOOCV for the top n = 5, 10, 15, 25 ranked candidates using our prioritization approach with single evidence type association scores and combined association scores. STRING assocation scores with a value less than the cut-off value were not taken into account. With a cut-off value of zero all associations contribute to the overall ranking score. Maximal TPR across all combinations of predefined values for the pI and Mw range is reported. For each value of n the highest TPR is indicated in bold.

ISSN: 1471-2105