Visualizing Molecular Structure / Property Correlations
Recently, I wrote to ask for suggestions in chemistry visualizations. Rajarshi Guha wrote in with several suggestions. One of the most interesting to me was to map results or numeric data to structures themselves:
In terms of mol vis itself, I’d like to see useful ways of mapping things to structures themselves - so fragment properties could be mapped to specific parts of a molecule, effects of descriptors in a QSAR model could be mapped to parts of a molecule and so on. Some of this has been done before, but is usually quite restricted.
I mentioned this idea as to colleagues “automatic annotation.” (After all, we can do what Rajarshi describes manually.) This relates directly to some of the work I did on my Ph.D. I ran computational predictions on electronic properties of molecules and drew out key effects via statistics.
One representation which was very useful to me was to indicate bond lengths in my molecules by single bonds and double bonds. In actuality, the results clearly indicated that many of the charged molecules I was considering had “unconventional” bonding.
I do think we need tools to allow us to quickly scan databases of molecules while mapping properties or statistical results on top.
Here’s a quick 2D mock-up:
Here I’m suggesting that we can map properties not just to an overall molecule itself, but also on top of particular sections of a molecule that we want to highlight. Honestly, I don’t think it’s all that difficult (famous last words), provided you have a flexible visualization tool.
Does this seem useful? Has it been done before?
4 Responses »
Comments
Subscribe to comments on this post via RSS or TrackBack
Rich Apodaca — 2 months, 3 weeks ago.
Geoff, I’ve been thinking along exactly the same lines. As you say, it’s not difficult, provided you have the right tool:
http://depth-first.com/articles/2008/02/06/the-chemically-enabled-user-interface-an-introduction-to-leafcutter
Rajarshi — 2 months, 3 weeks ago.
Hmm, I wonder how difficult it is. Say you have a set of substituted benzoic acids. And you’re interested in highlighting those COOH groups on which there is less charge compared to the average charge on that group over the whole dataset. If you had partial charges for te atoms this is not difficult at all.
In general, for atom based properties, I don’t think this would be very difficult. On the other hand making a generalized and intuitive interface to this would be challenging
This problem, is reminiscient of the ‘Grammar of Graphics’ (http://www.spss.com/research/wilkinson/TheGrammarOfGraphics/GOG.html)
Geoff — 2 months, 3 weeks ago.
Rich, obviously what you’ve been describing on your blog would offer ways of doing this in 2D. I’ve always been worried that 2D structure diagrams often leave out useful perspective. I think if you can do this, it would be a great tool.
Rajarshi, you seem to be interested in “fragment properties.” But if you have a molecular visualization tool (2D or 3D) which “lets you get under the hood,” you can imagine ways to do it. You point to the great examples in the statistical analysis world. But programs like SPSS or S or R or Matlab… these all have programming languages to let experts script almost anything they want to view. There are some examples of such programmable molecular visualization tools, but not necessarily flexible enough to do this.
Want to know how I’d implement your COOH group example? I’d script something which matches the COOH using SMARTS and sets the color of those atoms based on the numeric data.
In my Ph.D. dissertation, I changed bond orders to reflect actual bond distances, rather than the conventional “three-double bonds to a benzene ring” and “two double bonds to a thiophene.” You immediately saw changes in bonding structure — no need for a strange Excel plot.
No, I can’t do this yet using Avogadro (or anything else). But that’s why I’m posting challenges. This type of feature would clearly be helpful to chemists. Now we need to go try it. Why should we limit ourselves to the traditional chemistry representations?
Rich Apodaca — 2 months, 3 weeks ago.
Geoff, both 2D and 3D views are complementary. What works well for one usually doesn’t work well for the other.
BTW, JSpecView has some interesting demos set up:
http://wwwchem.uwimona.edu.jm/spectra/JSpecView/iranim/index.html
For example, this one lets you correlate IR spectral peaks with vibrational modes visualized by Jmol:
http://wwwchem.uwimona.edu.jm/spectra/JSpecView/iranim/oclanilineJ/oclaniline.html