Key points are not available for this paper at this time.
SUMMARY: Although the HMMER package is widely used to produce profile hidden Markov models (profile HMMs) for protein domains, it has been difficult to create a profile HMM for signal peptides. Here we describe an approach for building a complex model of eukaryotic signal peptides by the standard HMMER package. Signal peptide prediction with this model gives a 95.6% sensitivity and 95.7% specificity. AVAILABILITY: The profile HMM for signal peptides, data sets, and the scripts for analyzing data are available for non-commercial use at http://share.gene.com/.
Zhang et al. (Tue,) studied this question.