Alex's paper looks back at all of the fragment papers and looks at what was the starting point, where it went, etc. I took his data and applied my own filter to it (what one former colleague termed the ZOF, the Zartler Optical Filter).
The following two tables are from Chapter 2 of the upcoming book, on designing a fragment process.
This data is looking at the origin of the type of fragment strategy employed and the main method used to support it moving forward. Linking/assembly is exactly what it sounds like, attaching different fragments (zinc binding group and biphenyl a la Abbott). Anabolic is the logical growth of a fragment to a lead-like/drug-like molecule. Other could not be classified from Alex's paper.There were roughly the same number of projects that used either anabolic or linking. However, the anabolic method would appear to be better at maintaining or increasing ligand efficiency.
I also looked at NMR vs. X-ray vs. Other methods in terms of projects that maintain or increase LE. NMR and X-ray have the same number of projects (whose sum is equal to the total of Other). It does appear that NMR is better than X-ray in maintaining LE.
Here is the firebomb: With the well documented history of success, sufficient data to say that the anabolic approach is superior, and the NMR approach superior to X-ray, why is it that large pharma seems to be so slow in adopting what is an obviously successful method?
Could this mean that the average ligand efficiency of hits identified by X-ray screening is actually greater than that of hits identified by NMR screening?
ReplyDeleteThe tables only take into account *published* efforts. Unpublished approaches may still be more efficient.
ReplyDeleteRegarding NMR vs Xray:
Since biotech/pharma typically publishes work that isn't commercially viable, one could argue that the NMR approach leads the way in profitless endeavors.
On a more serious note - NMR will continue to struggle with larger protein targets. Xray isn't quite as size limited. Xray also gives more information for unexpected binding modes/protein conformations.
Interesting observations. Phil Hajduk and others have also noted that optimized molecules tend to be less ligand efficient than starting fragments, and in my own mini-analysis I saw similar statistics. In examples I looked at for a 2006 review (Curr. Opin. Biotech. 17 643-652), I found that, where ligand efficiencies were available for both starting fragments and final products, only 45% (5/11) of molecules derived from an “anabolic” or growth approach, and only 35% (6/17) of molecules derived from a linking approach had higher ligand efficiencies than one of the starting fragments. In the case of the linked compounds, none had higher ligand-efficiencies than both starting fragments.
ReplyDeleteHowever, we should keep in mind that, while ligand efficiency is a useful tool, it is not the ultimate goal. For example, in the case of Astex’s AT7519 (see 5 August post), the final drug had lower in-vitro potency and ligand efficiency than an earlier compound, but this was more than compensated for by an improvement in pharmaceutical properties.
One concern I have with the type of analysis summarised in the post is that a single starting point for further work is often assumed. I typically follow up hits with screening of close (e.g. substituent variation) and less close (pharmacophoric such as carboxyl to tetrazole) analogs. One advantage of fragment-based approaches is that you can often find the related structures because they are of lower molecular complexity. However, this approach to following up screens can obscure the identity of the original hit.
ReplyDeleteEver considered to use enzyme assays as a 1st line assay for fragment screening, then follow up with NMR or X-Ray AND SPR?
ReplyDeleteDoes the job!