Local Spatial Patterns (LSP)

LSP is a new protein comparison method developed at San Diego Supercomputer Center by Alexandr Kornev. It is based on comparing three-dimensional residue connectivity patterns. If two protein structures are analyzed, a residue connectivity map (graph) is built in each of them. Next, they are compared using such criteria as distances between residues and their mutual orientation. Once a matching subgraph (a set of trianglular residue connections) is found, it's added to the map.

Significance of the comparison graphs is in the fact that the most precisely positioned (and hence, likely to be functionally important) residues typically have more numerous connections with their neighbors. Therefore, a set of well-connected residues in a protein structure indicates the functional core that is composed of very accurately positioned amino acid residues. Number of connections at each residue is termed Involvement Score (IS).

Since the comparison is made between two proteins, the matching residue sets indicate a certain degree of conservation. This, in turn, may mean that residues with the highest involvement scores have a function that is conserved between the proteins being compared.

Comparison options

In order to detect similar connections within the two structures, not only geometric parameters are evaluated, but also the nature of the amino acids. How strict the filter, is determined by the option in the Residue similarity matrix panel. Selecting Identity matrix means that the only connections considered similar will be between exactly the same residues. For instance, only Ala-Leu, but not Ala-Val. Selecting Restrictive residue similarity matrix will allow some variation, so that hydrophobic residues will be treated as equal. If you choose Relaxed residue similarity matrix, the geometric shape of the side chains will also be used as a part of the criteria.

You can limit the extent of searched connections by specifying CA-CA distance cutoff. If any two residues are farther away than the cutoff, they will not be considered for a connection.

Important note: Selecting larger distance cutoff or relaxed similarity matrix will dramatically increase computation time. For large structures and relaxed options, the total time can exceed five minutes. This is due to the huge number of residue combinations that need to be found and analyzed before a comparison map is constructed.

 

Example:

Load two protein structures from the PDB. For example, 1HCK and 3LCK, which are protein kinases. First, run structure alignment using the corresponding option from the Structure menu. The two structures should overlay closely with minimal deviations, and the sequences will change accordingly.

To generate an LSP comparison map, select option Compare Local Spatial Patterns... from the Structure menu. At this point, you need to specify the two structures being compared. Select the first and second structure using the pulldown menus in the top panel of the dialog. Also select the appropriate chains that should be compared. In this case, select chains A in both structures. Which structure is selected first is not important. The rest of the settings should be left at their default values.

Once the comparison is completed, a small panel titled Local pattern clusters will appear. It will contain a list of clusters of connected residues that are similar in both proteins along with the number of residues in each cluster. Typically, the largest clusters are the most significant ones.

Select Cluster1 and click Show button and the bottom of the panel. The 3D structures will show connection graphs composed of red and blue lines. These indicate the connection graphs in each structure. The patterns are very similar, which indicates that the proteins are functionally related. And indeed, they are both protein kinases. Residues that correspond to each other in these two kinases can be displayed by choosing option Data->Show residue mapping in the small list panel. It will show the list of residues that are members of the cluster in both structures. This list, effectively, represents a functional alignment of the two kinases with the most significant residues included.

Distribution of Involvement scores can be visualized by clicking Score profile button in the cluster list panel. Another window will open, and show a bar graph of involvement scores plotted against the list of residues in the first protein entered in the initial dialog. This graph is a convenient way to assess importance of residue clusters in newly compared proteins.

Graph can be saved or printed. Moving the mouse over a bar will display the description of this specific entry in the top left corner of the window. The description includes full names of the corresponding residues in both proteins.