Structure Prediction: 1D Characteristics

1D Characteristics of a sequence are the ones that can be represented by a single value associated to each aminoacid (B. Rost). In explanation, the Secondary Structure (associated values: H -helix-, E -strand-, L -loop-, ...); accesibility (associated values: buried or exposed; or accesibility percentage); hydrophobicity, etc. Some of these characteristics are tabulated, as the hydrophobicity; others, as the secondary structure, can be predicted with a certain accuracy. The  1D characteristics of a sequence are very useful to the prediction of 3D structure.

Programs and Servers

• PredictProtein : The PredictProtein server is a powerful service for sequence analysis and structure prediction. Once a sequence is submitted, PredictProtein retrieves similar sequences in the database and predicts aspects of protein structure:

• Database searches:
1. generation of multiple sequence alignments (MaxHom)
2. detection of functional motifs (PROSITE)
3. detection of composition-bias (SEG)
4. detection of protein domains (PRODOM)
5. fold recognition by prediction-based threading (TOPITS)
• Predictions of:
1. secondary structure ( PHDsec, and PROFsec)
2. residue solvent accessibility (PHDacc, and PROFacc)
3. transmembrane helix location and topology (PHDhtm, PHDtopology)
4. protein globularity ( GLOBE)
5. coiled-coil regions ( COILS)
6. cysteine bonds ( CYSPRED)
• EXAMPLE OF PredictProtein SERVER OUTPUT.

Other topology features:

• TMHMM: Prediction of transmembrane helices in proteins (CBS, DK) [Output Example]
• COILS: Prediction of Coiled Coil Regions in Proteins (ISREC, CH) [Output Example]

THE SECONDARY STRUCTURE PREDICTION:

PredictProtein (PHDsec)	PHDsec predicts secondary structure from multiple sequence alignments. Secondary structure is predicted by a system of neural networks rating at an expected average accuracy > 72% for the three states helix, strand and loop (Rost & Sander, PNAS, 1993 , 90, 7558-7562; Rost & Sander, JMB, 1993 , 232, 584-599; and Rost & Sander, Proteins, 1994, 19, 55-72). Evaluated on the same data set, PHDsec is rated at ten percentage points higher three-state accuracy than methods using only single sequence information, and at more than six percentage points higher than, e.g., a method using alignment information based on statistics (Levin, Pascarella, Argos & Garnier, Prot. Engng., 6, 849-54, 1993).	[Output Example]
Jpred	Jpred is a internet web server that takes either a protein sequence or a mulitple alignment of protein sequences, and predicts secondary structure.  It works by combining a number of prediction methods to form a consensus.  For single sequences an automatic alignment method is used.  This alignment method takes your sequence  and scans against the non redundant sequence database. The hits are then post filtered with SCANPS, using length dependent cut-offs.  The resulting sequences are aligned  with CLUSTALW (v1.7).  The resulting alignments are automatically modified to have no gaps in the target sequence.  Then the prediction algorithms are run.  More detail can be found in the technical detail section.	[Output Example]
PsiPred	PSIPRED is a  secondary structure prediction method, incorporating two feed-forward neural networks which perform an analysis on output obtained from PSI-BLAST (Position Specific Iterated - BLAST) (Altschul et al., 1997). Using a stringent cross validation method to evaluate the method's performance, PSIPRED is capable of achieving an average Q3 score of nearly 77%. Version 2.0 of PSIPRED includes a new algorithm which averages the output from up to 4 separate neural  networks in the prediction process to further increase prediction accuracy.	[Output Example]