Hinge point predictions

Hinge point detection has been carried out to determine residues around which large protein movements are organized. Analysis has been performed by three different methodologies, each one with its own implementation. The different hinge point predictors have been coupled to both standard one and Gaussian RMSd fits, but the later is recommended.

BFactor slope change method

This method is based in the analysis of the B-factors landscape. The idea exploited is that a protein that has been re-centred using a Gaussian RMSd and has a hinge will display a fixed domain (low B-factors) and a floppy domain (large B-factors). This leads to a B-factors landscape dominated by low values in the fixed domain and very high values in the floppy domain. Hinge point is then located at the region of sharp slope change. To avoid discontinuities and reduce noise related to the roughness of the B-factor distribution the slopes were computed by averaging numerical values using different window sizes, checking in all cases for coherence (determined from the different windows estimates of the slopes).

The server labels as uphill those detections related to a change from low to high B-factors and downhill to the others.

The predictions are plotted as a histogram, with the downhill predictions in red and the uphill predictions in green. The aminoacidic sequence is drawn below the graphics, coloured in with the same code as these graphics. Brighter colours indicates increased confidence in the predicted hinge.

Force constant method

The method is based on the computation of a force constant for each residue, as stated in the paper Investigating the Local Flexibility of Functional Residues in Hemoproteins (Biophysical Journal 90:2706-2717 (2006)) by Sophie Sacquin-Mora and Richard Lavery.

The method computes a force constant for each residue that is dependent upon the distances between the residues along the trajectory:

Where is the distance between residue and residue , and refers to all the residues except , and .

As stated in the paper, the peak force constants will probably correspond to the residues in the interdomain region. This means that the hinge points are marked by the peaks in the landscape.

The force constant landscape is shown, and the top residues (top 20%) are highlighted in a residue sequence.

Dynamic domain detection method

This method relies on the proper clustering of the residues according to its correlation. It is based on the paper Probing Protein Mechanics: Residue-Level Properties and Their Use in Defining Domains (Biophysical Journal 87:1426-1435 (2004)) by Isabelle Navizet, Fabien Cailliez and Richard Lavery.

In this method, domains are detected by clustering the residues that maintaint their distances through the simulation. A kind of "correlation" matrix is computed using the distance differences between residues along the trajectory. This matrix have small values for correlated residues and high values for uncorrelated residues. This information is used to cluster the residues up to an empirical threshold value that produces a reasonable number of clusters.

Further refinement is performed on the clusters by checking the mean distance between the residues of each cluster and moving residues between clusters if this operation lowers the mean distance in the two affected clusters.

The hinge points are the contiguous residues found in the contact regions.

The result is presented as a coloured residue sequence, but also in the form of JMol applet with the backbone coloured according to the cluster assigned to each residue.