7

As part of some research I am doing, I want to do a head-to-head comparison of software for generating 3D atomic coordinates for molecules. Is there a list anywhere of chemical software that does this?

I was able to settle on the following constraints: freely available, or at least free for academic users; can be run on the command line (either Windows or *nix); and, can handle small to medium-sized molecules (I.e. I am not expecting the software to handle oligopeptides or other sizable biomolecules).

The input could be any of 1. a MOL file without 3D information; or 2. line notation like SMILES or InChI. I was envisioning doing tests for at least those formats, unless I am missing anything.

If there is no such list, would you be able to recommend any software for this task?

I am already using OpenBabel and RDKit to generate coordinates, so I got the idea to look for other software to compare with it.

(Survey papers that have previously done comparisons of this sort would also be welcome and appreciated, so that I can see what hasn't already been covered by previous researchers.)

Martin - マーチン
  • 44,013
  • 13
  • 159
  • 319
Joey Herrera-Orti
  • 143
  • 5
  • How "good" do the coordinates have to be? Do you want realistic bond lengths and angles only, or a sensible conformation as well? – Karsten Sep 03 '19 at 11:36
  • 1
    The "goodness"/"plausibility" of a structure's coordinates (I am still working on making this criterion more concrete and quantitative) would actually be my way of comparing/ranking various generators. (As I mentioned in my question, I want to do a head-to-head.) Certainly, the more "realistic" the generated structures are, the better a certain program is for at least that molecule. – Joey Herrera-Orti Sep 03 '19 at 11:42
  • 1
    This Wikipedia list should get you started: Cheminformatics toolkits. As far as I can tell, you have at least the dominant ones. For evaluating these structures, I suggest looking at CREST which is a conformer sampler based on the semiempirical GFN-xTB. I'd be very interested in your results. – Martin - マーチン Sep 03 '19 at 11:54
  • @Martin-マーチン - CREST is excellent and we use it, but it's much, much slower than these tools on standard organic molecules. – Geoff Hutchison Sep 05 '19 at 02:32
  • 1
    @Geoff I understand completely, and I agree, but I thought it's a good tool to address the "goodness" of a generated structure with some higher level of theory. Obviously, you can go further than that, too. – Martin - マーチン Sep 09 '19 at 08:29

2 Answers2

2

Summary: RDKit ETKDG is currently the best open source tool

This is an interesting question, since a reviewer just asked us to do such a benchmark.

The trick is that there's a fine distinction between "generating 3D coordinates" (i.e., one conformation) and "generating 3D coordinates and doing some conformer sampling" presumably with the latter helping to find a low-energy conformer, thermally-accessible ensemble, or a pose that might match an experimental structure or drug docking pose.

I'm not sure there are many papers on the first (i.e., generate one 3D structure - how good is it)? For one, it would depend a lot on your definition of 'good.' I personally prefer 'low energy conformer' because it's easy to define. Most people usually mean 'close to the experimental crystal structure' which is not usually the same as 'lowest energy'.

I digress. There are a few benchmark papers - largely grabbing a few conformers, usually including a mix of free or open source tools and commercial tools.

  1. Ebejer, J.-P.; Morris, G. M.; Deane, C. M. "Freely Available Conformer Generation Methods: How Good Are They?" J Chem Inf Model 2012, 52 (5), 1146–1158.
  2. Friedrich, N.-O.; Meyder, A.; de Bruyn Kops, C.; Sommer, K.; Flachsenberg, F.; Rarey, M.; Kirchmair, J. "High-Quality Dataset of Protein-Bound Ligand Conformations and Its Application to Benchmarking Conformer Ensemble Generators." J Chem Inf Model 2017 57 (3), 529-539.
  3. Nils-Ole Friedrich, Christina de Bruyn Kops, Florian Flachsenberg, Kai Sommer, Matthias Rarey, and Johannes Kirchmair. "Benchmarking Commercial Conformer Ensemble Generators." J Chem Inf Model 2017, 57 (11) , 2719-2728.
  4. Naruki Yoshikawa, Geoffrey R. Hutchison. "Fast, efficient fragment-based coordinate generation for Open Babel" J. Cheminf. 2019 11, art. 49.

While the newest method in Open Babel (development for v3) is faster than RDKit, particularly without using force field minimization, RDKit's ETKDG is very, very good - competitive with most commercial packages.

Geoff Hutchison
  • 27,740
  • 3
  • 79
  • 145
1

I hope I understand you correctly. For small size molecules, you can use the Avogadro program. For large systems (I use it with 40K atoms) molecules I would use VMD. If you want to generate better-looking figures I would use Chimera or PyMol which has a free version. Also, there is a very nice trial version of ChemCraft. If I remember correctly all the above-mentioned programs have both a Windows and a Nix version (at least Ubuntu compatible).

If you want to build or edit your current coordinates then ChemCraft is good though it doesn't handle large system so smoothly (although it should be said that I am visualizing on my laptop which does not have a dedicated GPU).

If you are looking for atomic structure 3D analysis like RMSD, bond angles, distances and etc. I myself use cpptraj in the AmberTools suite.

As Karsten already said, it is not very clear what exactly you are looking for in the software you are looking for.

Avishai Barnoy
  • 578
  • 2
  • 10
  • I am not at a computer right now to verify, but if all of your recommendations are usable on the command line, and can operate in batch mode, these would be quite nice. – Joey Herrera-Orti Sep 03 '19 at 11:37
  • VMD can be used from the command so is AmberTools cpptraj (trajectory analysis). I am not sure about Chimera, PyMol, ChemCraft or Avogadro. I used cpptraj to analyze large systems, 1us classical MD of 40K atoms size systems. – Avishai Barnoy Sep 03 '19 at 11:39
  • 3
    None of these tools generate 3D coordinates, they display them only, and you can edit them based on fragments and atoms. Avogadro uses Open Babel as an interface for 'creating' structures. – Martin - マーチン Sep 03 '19 at 11:42
  • If it's that 3D coordinates then Avogadro works, and if it's just with text I would use SMILES. Avogadro and VMD can be used to build molecules, so is ChemCraft. – Avishai Barnoy Sep 03 '19 at 11:44