Examples

The Figure below demonstrates the tendency of the SSGS algorithm towards secondary structure guided alignment: In part (a) of the Figure, we can view a case where a gap region was preferred over a mismatched region. In part( b) we see the alignment's tendency toward mismatches in non secondary structure-defined regions of the aligned target and model. The mismatched regions are backbone regions that are spatially separated.

A.

B.

 

in the Figure below we shows three examples of the results of the SSGS algorithm, when applied to protein fragments. All three cases show less than 20% sequence identity. In Figures (b) and (c) the two fragments derived from proteins of different super-families. In Figure (a) the two fragments are from proteins of the same family, but from different regions of the protein. This shows that the SSGS algorithm finds good structural matches even in cases where the sequence homology is very weak. As seen in the figures, the secondary structure match is preferred over matches of loops and unassigned areas even if as a result, the overall RMSD of the match is higher. The tendency of the SSGS algorithm to prefer secondary structure elements to loops coincides with the fact that loops tend to be less conserved structurally than secondary structures, even in closely related proteins. Therefore loops are likely to be less important for the structural conservation of the entire protein. A pair-wise matching algorithm that prefers secondary structure matching to loop matching is more likely to discover matches that are significant to the conservation and retention of the protein structure.

The results given by the SSGS algorithm differ from other non-secondary structure guided alignment tools, as the matching of structure assigned regions of the proteins is in general better in the SSGS method at the expense of loop area matching. An example of this preference can be seen in the figure below, where we can view two different alignments of the same two chains created by SSGS (a) and Multiprot (b). MultiProt is a protein structure alignment tool which aims to achieve the first two goals of SSGS, that is, to maximize the number of pairs of equivalent main chain atoms and to minimize the RMSD between the aligned atom pairs. The two protein chains fragments that were cut from Cyclin H (mcs2) from human, residues 126-174 and Phycocyanin beta subunit from cyanobacterium, chain L, residues 48-96. Both of these protein segments contain a helix-loop-helix assigned region. As can be seen in the figure, the SSGS resulting superposition exhibits an alignment in which the matching of the helix assigned residues was preferred over an alignment with more residues and a lower overall RMSD which does not fully align the helix assigned residues, as can be seen in the resulting superposition created by MultiProt. As can be seen in the Figure, the sequence alignment of the match obtained by the SSGS algorithm is better than the sequence alignment of the match obtained by MultiProt for the same model and target. This stems from the fact that the SSGS aligned the secondary structure elements better than MultiProt, at the expense of a poorer loop alignment. Evolutionarily, the loop regions are poorly conserved and exhibit a high rate of insertions and deletions. Thus, finding consensus residues might require a better alignment of the secondary structures that are more conserved, over aligning the loop assigned regions. This makes a secondary structure guided alignment more significant when attempting to find conserved regions and similar structural patterns among proteins from different families.

 

We compare the results of this algorithm with MASS (Dror O., 2003), an algorithm that performs a multiple structural alignment of proteins using secondary structure information. The figure below provides the alignment of two ptotein fragments using SSGS and MASS. The aligned segments are protein fragments cut from human Transglutaminase, chain A, residues 2-34 and Immunoglobulin from mouse, chain L, residues 516-549. Both protein segments contain a strand-loop-strand region. As can be seen, the structure region is somewhat better and the loop region is much better aligned using SSGS than using MASS. This stems from the fact that MASS first aligns secondary structure elements and later attempts to find the best alignment within this framework, while SSGS aligns all the residues in one stage, which may result in an alignment that prefers loop regions over secondary structure regions to maximize the overall score. In addition, a secondary-structure based algorithm such as MASS is likely to fail in finding a match in cases where a building block contains few or no secondary structure elements whereas SSGS, despite preferring the alignment of secondary structure elements, can still proceed. The structure guided sequence alignment shown in the figure shows that although the alignment scores are similar in the two methods, the SSGS guided alignment prefers one larger gap in the loop region over two smaller gaps as in the MASS case. Our tests indicate that SSGS is a good compromise between the two approaches represented by MultiProt and MASS.

 

 

* Dror, O., Benyamini, H., Nussinov, R. and Wolfson, H. Multiple structural alignment by secondary structures: algorithm and applications. Protein Sci, 2003, 12:2492 2507.

* Shatsky M, Fligelman ZY, Nussinov R, Wolfson HJ. Alignment of flexible protein structures. Proc Int Conf Intell Syst Mol Biol. 2000, 8:329-43.