DSIMB
For two pdbs with similari structures, but they generated different junctions
I have two similar protein structures.
I want to use SWORD2 to separete them into multiple regions and compare them region by region.
But you may see the following images that they generate different regions. Most of them are similar, but one of them has an extra region, "red" one.
I also draw the overlapped structures for the region in the same color. Most of them are similar
Could I use on of them as reference to "refine" the another one prediction? Thanks.
Hi, thanks for using our tool!
SWORD2 segments a single structure by optimizing compactness and sparsity on its contact graph. Two very similar structures can receive slightly different partitions, especially around flexible termini or hinge loops. In your example, the extra red region is likely a small, weakly connected extension that crosses the acceptance threshold for one structure but not the other.
SWORD2 does not currently support a reference-guided mode that constrains the segmentation of structure B using boundaries from structure A.
The SWORD2 is an excellent software and it uses an clever idea to segment the protein structure.
I can understand the extra thing before. But here are two examples I also meet some issues. Of course, they are only some exception and not seen in a lot of proteins, but they do annoy me in such cases.
You can see that they are also two very similar proteins. What I expected is that I can use the "junction" to compare them piece by piece. But as you can see above, there are slightly shifting and make me hard to align their pieces well.
Do you have any solution for adjusting them and make them more consistent to each other?
Here are two pdb for your check.
Did you use SWORD2 online at https://dsimb.inserm.fr/SWORD2 ? It looks a bit too much segmented in your figures. The alternative n°1 looks better to me t1: https://www.dsimb.inserm.fr/SWORD2/TMP/04_09_2025_88498/results.html and t2: https://www.dsimb.inserm.fr/SWORD2/TMP/04_09_2025_42043/results.html
For me, I use the offline version. But I also tried online server for validation.
The large number of colors are because I use all the "Junctions" to separte the protein instead of using those "Alternative Partitioning". So maybe it will introduce more partition compared to "Alternative Partitioning"
For alternative n°1 , it looks better and less colors. But it still have the shifting problems.
I assume that the "red part" in the left figure should be shifted to the right terminal a little bit. And then these two are consistent.
Yes true, I saw that afterwards sorry. It looks like the algo is impacted somehow by the difference in secondary structures between 70-80 residues, the coil is a bit different.
@jcgelly do you have a clue on this?