I have a series of alanine dipeptide trajectories and would like to get the conformation type ($\alpha$, $\beta$, etc.) from the atom coordinates or dihedral angles. Does any package (like RDKit for instance) answer this?

  • 2
    $\begingroup$ +1. Thank you for splitting the question into two! You can even link the other question in this question if you'd like to do that. $\endgroup$ – Nike Dattani Sep 2 '20 at 17:49

[1] is a nice article that compares a couple of methods of protein structure assignment based on how well they agree with each other and how well they agree with expert assignments made in the Protein Data Bank. In the paper, they develop a new program called KAKSI, for which they make include an install link at the end of the paper.

It's important to note that assigning secondary structure is somewhat subjective, so different programs will have different criteria for what constitutes an $\alpha$ helix or $\beta$ sheet, both in terms of what features are considered (carbon-carbon distances, $\psi$/$\phi$, hydrogen bond patterns) and what values correspond to a particular secondary structure element.


  1. Taly, J.-F.; Martin, J.; Letellier, G.; de Brevern, A. G.; Gibrat, J.-F.; Marin, A. Protein secondary structure assignment revisited: a detailed analysis of different assignment methods. BMC Struct. Biol.2005,5, 17.

Alanine dipeptide (AD) is a very small system with only 3 residues (2 of which are capping residues). If you're looking for $\alpha$ helix or $\beta$ sheet structures, they do not exist in such a small system. The $\alpha$ helix, for example, is generally defined by the existence of hydrogen bonds between backbone atoms of residues $i$ and $i+4$. Many software packages won't assign any labels to such small systems, since secondary structure requires a little more context.

There are a few different terms for the conformational states of AD: you'll see $\alpha_R$, $C_{7eq}$, $C_{7ax}$, $\alpha_L$, etc. There are challenges with creating general software to label these kinds of conformational states. First, the software would need to know whether your simulations were gas phase or solvated. This has a large effect on the stable states of AD. The locations of the minima also change depending on the force field. If I recall correctly, $\psi=0$; $\phi\approx-100$ degrees is near the free energy maximum of a transition barrier in AMBER96, but near a stable state (free energy minimum) in AMBER99SB-ILDN! So without more information about your simulation, it would be impossible for any software to determine the states.

As a practical matter, the way I usually go about labeling states in such a system is to write a small bit of Python. Tools like MDTraj, pytraj, and MDAnalysis make it easy to extract the Ramachandran angles. For AD, you can probably figure out state boundaries by eye after plotting in the Ramachandran plane (or look at an existing free energy diagram for your force field, since AD is so well-studied).


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.