Protein Structure Prediction

Method

• Secondary Structure Prediction
• Solvent accessibility prediction & Folds
• Transmembrane helix and signal peptide prediction
• Prediction of coiled-coils
• Prediction of O-glycosylation sites
• Homology modelling
• Threading

---

Reviews

EMBL list of sites

collection of prediction services: Overview and links to services for predicting secondary structure, solvent accessibility, homology modelling and threading (MRC, Cambdridge, England).

PredictProtein: Multiple sequence alignment (MAXHOM); prediction of secondary structure (PHDsec), solvent accessibility (PHDacc), transmembrane helices (PHDhtm), transmembrane topology (PHDtopology); and threading (PHDthreader).

NIH molecular Modelling Homepage A modelling homepage with links

Secondary Structure Prediction

Name	Accuracy	Comments
PHDsec: http://www.embl-heidelberg.de/predictprotein/	> 72% (+/-10%, one standard deviation)	Multiple alignment-based neural network system.
NSSP: http://dot.imgen.bcm.tmc.edu:9331/pssprediction/pssp.html	> 71%. Evaluated on > 200 unique proteins.	Multiple alignment-based nearest-neigbour method.
SOPM: http://www.ibcp.fr/predict.html	> 70%.	Multiple alignment-based method combining various other prediction programs.
DSC: http://bonsai.lif.icnet.uk/bmm/dsc/dsc_read_align.html	70%	Multiple alignment-based program using statistics.
SSPRED: http://www.embl-heidelberg.de/sspred/ssp_mul.html	> 70%.	Multiple alignment-based program using statistics.
MultiPredict: http://kestrel.ludwig.ucl.ac.uk/zpred.html	> 65%	Multiple alignment-based method using physicochemical information from a set of aligned sequences and statistical secondary structure decision constants.
PSA: http://bmerc-www.bu.edu/psa/		The PSA server analyzes amino acid sequences to predict secondary structures and folding classes.
NNPREDICT: http://www.cmpharm.ucsf.edu/~nomi/nnpredict.html	> 65%	Single-sequence based neural network prediction.

Solvent accessibility prediction

Name	Accuracy	Comments
PHDacc: http://www.embl-heidelberg.de/predictprotein/	> 75% (+/-10%, one standard deviation),	Multiple alignment-based neural network system.
Fold-recognition		UCLA-DOE STRUCTURE PREDICTION SERVER

Transmembrane helix and signal peptide prediction

Name	Accuracy	Comments
PHDhtm: http://www.embl-heidelberg.de/predictprotein/	> 95% (+/-10%, one standard deviation),	Multiple alignment-based neural network system predicting the locations of transmembrane helices
TMAP: http://www.embl-heidelberg.de/tmap/tmap_sin.html	> 95%.	Single sequence-based statistical prediction of the locations of transmembrane helices.
PHDtopology: http://www.embl-heidelberg.de/predictprotein/	> 85% of all proteins all helices and topology are predicted correctly	Refinement of PHDhtm by dynamic programming and prediction of topology (orientation of N-term with respect to membrane).
TMpred: http://ulrec3.unil.ch/software/TMPRED_form.html		Single sequence-based prediction of location and topology for helical transmembrane proteins using statistics and similarity metrices.
Signalp : http://www.cbs.dtu.dk/services/SignalP/		Neural network prediction of presence and location of signal peptide cleavage sites in amino acid sequences from different organisms: Gram-positive and Gram-negative prokaryotes, and eukaryotes.
DAS - Dense Alignment Surface prediction of transmembrane regions in proteins

Prediction of coiled-coils

Name	Accuracy	Comments
COILS: http://ulrec3.unil.ch/software/COILS_form.html		Single sequence-based prediction for coiled-coil regions using statistical patterns of coiled-coil proteins in the database.
Paircoil		predicts the location of coiled-coil regions in amino acid sequences.

Prediction of O-glycosylation sites

Name	Accuracy	Comments
NetOglyc : http://www.cbs.dtu.dk/netOglyc/cbsnetOglyc.html		Neural network predictions of mucin type O-glycosylation sites in mammalian proteins.

Homology modelling

Name	Accuracy	Comments
SWISS-MODEL: http://expasy.hcuge.ch/swissmod/SWISS-MODEL.html		An automated knowledge-based protein modelling server ; first approach and optimise (Peitsch M.C. Protein Modelling by E-mail. Bio/Technology 13:658-660. (1995)

Threading

Name	Accuracy	Comments
PHDthreader: http://www.embl-heidelberg.de/predictprotein/	< 30%, less than 30% of the predicted first hits are true remote homologues. Evaluated by cross-validation on 89 unique protein structures.	Prediction-based threading detecting the fold type and aligning a protein of unknown structure and a protein of known structure for low levels of sequence identity ( < 25%).
T3P2: http://www.mbi.ucla.edu/people/frsvr/frsvr.html		Prediction-based threading detecting the fold type and aligning a protein of unknown structure and a protein of known structure for low levels of sequence identity ( < 25%).

  Protein Structural Analysis, BMERC. Protein Structural Analysis (under construction)