I am trying to find a way of calculating the 3 principal moment of inertia of a molecule, sorting the value in ascending order and dividing the two lower values by the biggest value. In this way I can have a descriptor that can be plotted in triangle plots . They can describe the shape of a molecule or 3-Dimensionality. Is this something that Knime already have in the set of descriptors? Or can be done easily?
The MOE nodes make pmi descriptors available in KNIME - but, of course, you need access to an installation of MOE to use them.
If you do have access to MOE, then the nice thing about these descriptors is that you save yourself a couple of (trivial) steps - because they are already ordered, and there are also pre-calculated ratios of the two smaller / biggest (npr1, npr2)!
I did something like this a while ago in KNIME - but was also trying to get a sense of conformational flexibility for the molecules, as well as shape (I guess I was actually after a shape flexibility descriptor - by looking at the spread of conformers in the npr1 and npr2 dimensions).
I have modified my workflow by including a random set of DrugBank small molecules as the input. The 'triangle' plot is achieved with the R View (Remote) node - having installed ggplot2. This was my first go at using R - so apologies for the very clumsy script (it sort of implements a manual 2D box-plot; instead of doing 2D variability 'properly' - I'll come back to it one day...)!
I have attached the workflow (data included), and a picture to give an idea of the sort of output.
Hope this gives some ideas / is of some use.
EDIT - I don't seem to be able to get the pic to stick! The link is:
Hopefully someone else can point you in the right direction for an alternative way to calculate the pmi descriptors. I realised that the picture link i had posted was not correct - I have edited the URL in my original reply - hopefully it works now(?)
Actually in knime node the moment of inertia is not present, you can use CDKgui from Rajarshi Guha or PaDEL descriptor gui these are opensource but external to knime.
I guess you're referring to W. H. B. Sauer and M. K. Schwarz in J.
Chem. Inf. Comput. Sci., 2003, 43, 987-1003.
As James already mentioned the MOE descriptors can be used to do such an analysis in KNIME. If you have no access to MOE contact support@chemcomp.com for an evaluation copy.
I had pretty-much forgotten all about posting this workflow (it was over 5 years ago!!) and your question encouraged me to relook at the task in light of changes to KNIME between then and now.
I have reworked the workflow with the following key points:
The PMI calculations are now handled by an internal Vernalis PMI node
The conformer generation is handled by RDKit nodes
The R script now runs in the KNIME R View (table) node - no longer requiring an R Server
The points on the triangle plot are now coloured by energy instead of fraction polar surface area (this last change is partly because it could be useful, and partly because I am not aware of a free way of generating this descriptor)
As stated above, the PMI node is an internal Vernalis node that we have developed. We have a good track record of getting our internal nodes approved for release externally - so I will if we can get this into the nightly build asap.
If/when this happens I will post back here with an update, and a copy of the workflow (as it won't be very useful to you right now...).
Kind regards
James
PS As already mentioned - you can, of course, contact CCG regarding the MOE nodes and make use of the existing workflow right now
Just to add to James' comment above, there is now a PMI Calculation node in the Vernalis plugin - currently only in the nightly build. If you update to version 1.6.4 of our plugin on the nightly build then you should get the nodes.