06 October 2010

Commercial fragments – how do they compare?

Earlier this year we updated our list of commercial fragment suppliers. Now Chris Swain at Cambridge MedChem Consulting has analyzed the structures and properties of eleven of these.

A major conclusion, which will be a disappointment to purchasers of fragments but a boon to suppliers, is that there is very little overlap in terms of exact molecules. Given the vastness of chemical space this shouldn’t be too much of a surprise, but it is striking that, of the 40,000+ molecules represented, less than a dozen are sold by four or more companies. Looking at molecular similarity rather than identity increases the amount of overlap, but for the most part each collection is quite differentiated from the others.

Where it gets really interesting is in the analysis of chemical properties, which Swain has calculated for each fragment set. These include cLogP, molecular weight, polar surface area, H-bond donors and acceptors, heavy atom count, and rotatable bond count. Here the collections are dramatically different, with some being strictly Rule-of-3 compliant while others are much less so. There are also interesting differences in distributions: some collections are distributed around a low molecular weight median, while others are biased towards larger molecules. Finally, an analysis of molecular diversity reveals some collections to be very diverse while others have clusters of closely related molecules.

Of course, people differ in how much weight to put on simple molecular properties. Also, an analysis such as this is necessarily a static snapshot: commercial offerings change over time, and new suppliers continue to enter the market. Moreover, some of the companies offer many more fragment-sized molecules as extension sets beyond their core fragment collections. Still, this is a valuable resource for anyone building or expanding a custom fragment collection. The only thing that would make it even more useful would be price per fragment!


Lionel C. said...

I would like to notice that NO standardization have been applied before the analysis, only format converstion by Babel.
A lot of problems can come from NO standardization before comparing chemical libraries....

Chris said...

I actually used the "Wash" command in MOE which should help.
Should have mentioned it on the page.

Peter Kenny said...

Rule of 3 compliance is something that always amuses me. Vendors tout the 'Ruleof3-ness' of their offerings but nobody (including the Rule's original proposers) actually say how the hydrogen bond acceptors or donors are defined. Will a hydantoin violate the Rule of 3? While I can't answer this question, I would be happy to have hydantoins in a fragment screening library.

Another point worth raising is that the presence of ionisable groups never seems to make it into analyses like this one. The presence of a functional group that is ionised under assay conditions will typically favor solubility. Have a look at this article to see how we used knowledge of ionisation in library design:


Chris said...

I actually calculated the most acidic and basic pka for each molecule using the ChemAxon tools, in addition to LogD.
I've now categorized all the fragments into acid, basic, neutral and zwitterions and added it to the profiles.

Chris said...

Several companies have sent me updated listings so I've redone the analysis. What else are weekends for!

Chris said...

Several new companies have sent me their collections to add to the comparison. So I've updated the page with the current collections.