Search prediction results

---

CRC64 (e.g. AB01C6346CB7FE64):
Uniprot (e.g. P78310):

---

The PONGO system has been described in:

• Amico M, Finelli M, Rossi I, Zauli A, Elofsson A, Viklund H, von Heijne G, Jones D, Krogh A, Fariselli P, Martelli PL, Casadio R - PONGO: a web server for multiple predictions of all-alpha transmembrane proteins - Nucleic Acids Res 34 (Web server issue):169-172 (2006)

---

The annotation server consists of six programs for the transmembrane annotation of the human proteome and three methods for the detection of three major post translational modifications, namely the cleavage of signal peptide, the GPI-anchoring, and the disulfide bonding.

TRANSMEMBRANE ANNOTATION

1. TMHMM2.0 s a predictor of transmembrane helices in proteins based on hidden Markov models developed by A. Krogh, B. Larsson, G. von Heijne, and E. L. L. Sonnhammer (Journal of Molecular Biology, 305(3):567-580, January 2001). TMHMM2.0 has the great advantage of beeing very fast, being based on only single sequence information.
2. Polyphobius developed by L. Kall, A. Krogh and E. Sonnhammer (Bioinformatics, 21:I251-I257, 2005) is based on a hidden Markov model that predicts transmembrane topology and signal peptides together. The decoding algorithm combines the features computed by the hidden Markov model on a set of homologous sequences by considering the average of the posterior label probabilities of each position in a global sequence alignment. The algorithm is an extension of the previously described optimal accuray decoder, allowing homology information to be used.
3. MEMSAT3 is the new version of the predictor of transmembrane helices in proteins developed by D.T. Jones (Bioinformatics 23:538-544 2007). This new version exploits the evolutionary information derived by multiple sequence alignment by means of Support Vector Machines.
4. ENSEMBLE2.0 is similar to ENSEMBLE, but it is endowed with a new method to assign topology (unpublished) that overcomes the prediction of topology by our consensus method, previously described (Martelli et al. Bioinformatics 19:I205-I21, 2003)). The algorithm used to assign the topology is the Posterior-Viterbi (Fariselli P, Martelli PL, Casadio R. BMC Bioinformatics 6 Suppl 4:S12, 2005). The models considered by the Posterior-Viterbi algorithm are the amphipatic and hydrophilic HMMs described in (Martelli PL, Fariselli P, Casadio R Bioinformatics Suppl 1:I205-I211, 2003)
5. PRODIV_TMHMM_0.92 is a new version of the a predictor of transmembrane helices in proteins developed by H. Viklund and A. Elofsson (Protein Sci. 13:1908-1917, 2004) which uses a hidden Markov model similar to TMHMM, but exploits the evolutionary information derived by multiple sequence alignment.
6. CINTHIA is a Consensus of INternational Transmembrane Helical Intelligent Annotators (CINTHIA) that is a non-trained metapredictor that exploits the predictions of made by ENSEMBLE2.0, MEMSAT3 and PRODIV_0.92. The best scoring model is found by means of the MaxSubSeq algorithm (Fariselli et al., Bioinformatics 19:500-505, 2003). CINTHIA outperforms all the three considered methods.

POST TRANSLATIONAL MODIFICATION ANNOTATION

1. SPEP is a predictor for the N-terminal signal peptide, based on Neural Networks (Fariselli P, Finocchiaro G, Casadio R, Bioinformatics 19:2498-2499, 2003). Since most of N-terminal signal peptides of immature proteins are mispredicted as trasmembrane helices, before running any of the predictors for transmembrane annotation (but Polyphobius) we test the presence of the signal peptide. In case of a positive answer, we cut the corresponding predicted segment. Sequences processed with Polyphobius were not cut, since this methods integates a signal peptide predictor.
2. PredGPI is method for predicting the presence of a GPI-anchor, based on Support Vector Machines and Hidden Markov Models (Pierleoni A, Martelli PL, Casadio R, submitted, 2008). Since most of C-terminal cleaved peptides of immature proteins are mispredicted as trasmembrane helices, before running any of the predictors for transmembrane annotation we run PredGPI. In case of a positive answer, we cut the corresponding predicted cleaved segment.
3. CYSPRED2 is a new version of the predictor of disulfide-bonded cysteines with a hidden neural network by Martelli PL, Fariselli P. and Casadio R. (Proteomics 4:1665-1671 2004).