I recently updated the fragment collections page this included updating the physicochemical property profiles adding npmi (Normalized ratio of principle moments of inertia) as described by Sauer WH, Schwarz MK (2003) Molecular shape diversity of combinatorial libraries: A prerequisite for broad bioactivity. J Chem Inf Comput Sci 43:987–10030. DOI As the image below shows this gives a view of the shape of the molecules as to whether they are rod, disk or sphere like.

Whilst this works very well for individual compounds or small libraries the plot becomes a blur of overlapping points for larger collections and it is not really possible to compare collections. Whilst it may be possible to generate a single number as the “average” of each collection I’m not sure how useful it would be. So with help from Matt I decided to divide the plot into three sections as shown below.

The centre point (c1, c2) was calculated using ( 0.5, (2sqrt(0.5) + 0.5)/(2sqrt(0.5) + 1) ) which is about (0.5, 0.793).

Each of the points in the plot was then assigned to a category using:

If a point is below both lines then: (0.5 - 0.25) * (npmi2 - 0.75) - (0.793 - 0.75) * (npmi1 - 0.25) < 0 and (0.5 - 0.75) * (npmi2 - 0.75) - (0.793 - 0.75) * (npmi1 - 0.75) > 0 then it is disc-like.

If not, then it is rod-like if npmi1 < 0.5 and sphere-like if npmi1 > 0.5.

The result for a 35,000 compound collection are shown below, with the points colour-coded by the assigned category.

We can then create a categorical plot as shown below.

I plan to update all the physicochemical profiles of all the fragment collections next week.