The impact of SignalP 4.0 on the prediction of secreted proteins

Secreted proteins play important roles in eukaryotes across each of the kingdoms of fungi, animals, plants and protists. The majority of secreted proteins are signal peptide dependent proteins so their prediction relies on the SignalP family of predictors for signal peptides. The recent release of SignalP 4.0 prompts the question of its impact on the prediction of secreted proteins when compared to current approaches utilising SignalP 3.0. We performed comparisons of SignalP 4.0 and SignalP 3.0 when used alone or in combination with other tools. We used a large dataset for each of the four kingdoms. Results are reported for sensitivity, specificity, and Matthews Correlation Coefficient (MCC) in each case. In terms of MCC, the best performance of any combination of tools always involved SignalP 4.0. In particular for the plant and protist kingdoms there were notable gains due to the use of SignalP 4.0.