what is the threshold RMSD value for determining protein structure similarity

Question

[smjsardar]

I predicted mutated protein structure (via MODELLER) from a reference protein structure. I then compared these mutant structures with my reference by superimposing and finding the RMSD values. Now what value of RMSD would determine if my mutated protein structures are different from my reference? Can I say values above 0.5 Angstrom means there is a significant difference between the two structures?

Answer 1

[Joe]

There's no magic number. An RMSD of less than about 2 Å would generally be considered very close, but there's no absolute rule. A single mutation could have a very profound effect on structure, or could make essentially no difference depending on where it is in the structure.

You need to think about your actual biological question.

Answer 2

[jgreener]

To add to the above answer, RMSD is dependent on the size of the protein.

If you are looking for a cut-off score, with all the caveats that entails, I would recommend using TM-score (or the related TM-align). TM-score runs from 0 to 1 and a TM-score of 0.5 or higher indicates that the two structures probably have the same fold.

Check out the CASP website and papers for more on this, they think pretty hard about this problem.