This server is designed to predict local protein structures and protein flexibility from its sequence. Results can be visualised at the amino acid level through a table and graphics.

Protein Local Structure Prediction

It is now admitted that the folded state of proteins, that is, the native 3D structure, can be described by a limited set of recurring local structures (Fitzkee et al., Trends Biochem. Sci. 2005). This observation led to the development of fragment libraries designed to characterize in the most suitable way, the local structures of all proteins with known 3D structures. These libraries consist in a finite set of representative structural fragments. Nowadays, when no homologue protein is available, the most successful methods for predicting global 3D protein structures use fragment assembly techniques.

A library of 120 3D structural prototypes encompassing all known local protein structures has been developed (Benros et al., Proteins, 2006). These Local Structure Prototypes (LSPs) were mean representative fragments of 120 overlapping structural classes of 11-residues fragments. They ensured a good quality of approximation. An associated local structure prediction method from sequence was also created. Its principal interest was to propose a limited number of relevant structural candidates for a given target sequence.
Recently, we achieved a balanced improvement of the prediction rate by coupling evolutionary information with support vector machines (SVMs). A very satisfying correct prediction rate of 63.1% was obtained for 5 proposed candidates (Bornot et al., Proteins, 2009). This prediction method is implemented in this web service.

Protein Flexibility Prediction

In the same way, protein structures are not rigid macromolecules. We analysed local structure flexibility features in proteins by relying on: (i) B-factors from X-ray experiments and (ii) backbone fluctuations in solution observed in molecular dynamics simulations. Finally, an original flexibility prediction method from sequence was developed (Bornot et al., Proteins, 2011). Three classes of flexibility are considered. Very few confusion between rigid and flexible classes was observed. Only 13.5% of rigid residues were predicted as flexible and reciprocally, only 5.8% of flexible ones were predicted as rigid.This method is implemented for this web service.

Launch a prediction

Paste your sequence file (fasta file format):


Browse your sequence file (fasta file format):

Note that data older than 48 hours are deleted.


Alexandre de Brevern

Aurelie Bornot

Pierrick Craveur

Catherine Etchebest

Jean-Christophe Gelly