I'm wondering about the minimal set of parameters necessary to define a protein's structure. My understanding is that the backbone geometry is defined by the phi and psi angles (torsion angles), and then from this template the side-chain rotamers are defined. Is this sufficient to yield a geometric structure, or are there other factors involved?

On a related note, how exactly are two protein structures compared? I often see references to percent similarity, but what does this really mean? In the Protein Data Bank, atomic coordinates are given, but my intuition is that coordinates are only important relative to all the other amino acids in the chain. In other words, absolute coordinate comparison would seem to be uninformative.

  • $\begingroup$ If you're going for bare minimum data length, you'd also have to consider commonly occuring structures - think Huffman coding. Alpha helices and beta pleated sheets, for example, are a lot of data that can probably be compressed quite readily. It also occurs to me that one could define very rough parameters which give many possible configurations, but only one stable one, amongst other shortcuts. $\endgroup$ – Watercleave Feb 16 '15 at 17:48
  • $\begingroup$ Ah, interesting point I hadn't considered (encoding). I think I'm going more for geometric parameters. Perhaps another way of saying it - are torsion angles and side chain rotamers sufficient to completely determine a protein's structure? $\endgroup$ – ZenPylon Feb 16 '15 at 17:54
  • $\begingroup$ By "protein structure" do you mean the position of the atoms in a protein at a particular point in time? $\endgroup$ – Lubo Antonov Feb 17 '15 at 10:48
  • $\begingroup$ @LuboAntonov Yes, exactly - specifically the folded state. $\endgroup$ – ZenPylon Feb 17 '15 at 13:03
  • $\begingroup$ Your percent similarity is sequence identity. $\endgroup$ – Mithoron Feb 19 '15 at 1:15

It depends on how accurate you want the description to be. Ideal bond lengths are normally assumed, so the structure can be described using φ, ψ, ω (for the backbone) and χ1-χ5 (for the sidechain) angles. This is what is used in most cases. However, the positions of the hydrogens in the sidechains are not uniquely defined by this description. This also does not describe hydrogen bonds and disulfide bridges, but those could be deduced from the relative positions.

You can get more information here: http://kinemage.biochem.duke.edu/teaching/anatax/html/anatax.1b.html

This description should be enough for most applications, though.


Comparing Protein Structures

There are a number of things you can look at such as root mean square deviation. This method compares the co-ordinates of two structures by superposing them on top of each other.

  • RMSD: This is implemented in the modelling Modeller or can be used in USCF Chimera
    • RMSD will provide a number in angstroms squared, with higher values indicating structural differences. A RMSD of 0.0 would indicate the structures are identical. There are different types of RMSD calculations such C-alpha, backbone and all-atom. C-alpha RMSD will only look at the C-alpha atoms to calculate the RMSD. Backbone RMSD will look at the backbone only and all-atom will take into account all the atoms to calculate the RMSD.

Other Things to Look at

  • Assuming the structure was created yourself, was the structure minimised? To remove clashes?

  • In addition phi, psi angles and mentioned above have you checked for clashes in your structure? Clashes can be checked in USCF chimera which checks for clashes with a default value of 0.6 Angstrom van der Wall radii overlap.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.