rmsd

[Brian the Roman]

Yet another stupid question...

What does the RMSD really measure? The unit is Angstroms so it's probably a distance of some sort. Or, it could be the wavelength of light...

I ask because I've been looking at the graphs of the top guys and I've noticed that crease energy and rmsd (which should both roughly rise and fall together, I believe) are sometimes significantly out of synch as they are in Dano's structures.

Does this suggest that the lowest energy is not always the best, or that one or both of our algorithms for calculating the crease energy and rmsd are flawed?:confused:

ms

[bwkaz]

RMS is root-mean-square, and D here is deviation.

So I would assume that Howard is calculating it by running through each amino acid, measuring how far "off" it is from the real structure, squaring that distance, and adding them all up. Then he divides by the number of AA's (to get the mean square deviation) and takes the square root.

So it's a measure of how far "off" the structure is from the real protein, which is why you need to know the real protein structure to be able to calculate it.

As to why crease energy isn't always quite synchronized, that's probably because the correlation coefficient between the two isn't exactly 1. But I'd guess it's 0.97 at least, which is still pretty good.

One other thing is that you don't need the exact (0.0 RMSD) structure to create (for example) drugs to target a specific protein. IIRC, Howard has said that RMSD values of less than 6 (or so, depending on the protein length) are pretty much good enough. So we don't need perfect scoring functions (crease energy is one scoring function), just a couple of decent and fast ones, and one good but slower one (that Howard, et al. can run on the best few of the submissions later).

[Brian the Fist]

Yup, thats more or less right. To design drugs reliably however, you need 3A or better typically. What I did say is under 6A it is usually good enough to recognize the general fold and topology of the protein.

And all energy functions are flawed and approximate, hence the discrepancy. Our search for the 'perfect' scoring function is an ongoing one...

[bwkaz]

Ah, I see. Thanks for the clarification. :thumbs: