04 January 2012

19F NMR...Just Like 1H NMR

Fluorine is a powerful nucleus for NMR: it has almost the same sensitivity as 1H and is 100% abundant. 19F NMR based screening is relatively new in the pharmaceutical industry. One of the biggest proponents of this approach has been Claudio Dalvit. In 2009, Dalvit and colleagues (J. Am. Chem. Soc. (2009) 131:12949) proposed the LEF (Local Environment of Fluorine) approach which allows for efficient screening of even weak binding fragments containing fluorine and the discovery of fluorophilic protein environments on the target. Despite the obvious advantages of 19F, it still has not been taken up widely in the industry. A new paper in J. Med. Chem. from Amgen may change that.

In this paper, they demonstrate multiple practical applications of 19F in FBDD, essentially demonstrating congruence for 19F-based methods with 1H-based methods. This is a particularly important as most people are conversant with how to understand and utilize 1H-based data; there is no "re-learning" necessary to adopt 19F based data. One potential obstacle to adoption is the argument that the 19F fragments would not be "fragment-y" or would be unlike other libraries and thus have little cross-fertilization. As shown in the table below, their 19F Fragment library is based on typical Ro3 rules with similar size and diversity.

As can be seen here, this library is relatively diverse. The authors argue that ~30% of "Fragment Space" is covered by this library. But, due to congruency with 1H molecules, this is more than sufficient to give adequate screening hits.The paper demonstrates the utility of this approach on BACE (one of everybody's favorite targets). In less than 24 hours they were able to screen the entire library and identified five compounds (out of 6 hits) for follow-up. They introduce a method for Kd determination by NMR which cannot be done by 1H-based methods, differential Chemical Shift Perturbation: which utilizes changes in the 19F chemical shifts to determine the Kd. In cases where chemical shifts are not observed, the authors state you can use differences in line broadening as the measurement. This is an interesting and novel approach to Kd determination and they validated this approach by using SPR in parallel. The authors also demonstrate that the Holy Grail of FBDD (simultaneous binding fragments) can be identified and oriented based upon 19F-19F iLOE, exactly analogous to 1H-1H NOEs.

This is an excellent paper, that I hope sparks more people to investigate the utility of 19F-based screening and post-screening confirmation.



4 comments:

Dan Erlanson said...

This is an interesting approach, and I’ve always wondered why it hasn’t been used (or at least reported) more often. I know at least one vendor of commercial fragments offers fluorine-containing fragments specifically.

Darren Begley said...

I've always liked this approach, with evidence of its utility dating back to the 1970s (see Bird et al, JACS vol 100 p7478, 1978). Importantly, it expands the range of binding affinites for small molecules in ligand-observe mode, compared to other proton-based methods.

I suppose one reason this approach is not more common is instrumentation. Traditionally, most systems are configured with either HCN or H-X probes, with an X-nucleus amplifier which doesn't go to 19F. And those of us with cryoprobes are loath to swap them out frequently.

As far as the library goes - I think if researchers scoured their pre-existing fragment libraries, they might find they already have enough fluorinated compounds to do a fragment screen.

mostafa said...


Has anyone else actually validated the dCSP method in this paper? I can't get my head round how they have linearised the relationship between CSP (or dCSP) with ligand concentration?

Anonymous said...

The 19F chemical shift changes in Figure 4 were 8.3Hz (0.018ppm at 500MHz NMR spectrometer) and 6.9 Hz (0.015ppm). (really small numbers to me. is this often the case in 19F NMR? and is this practical? )

Using equation-4 I got a Kd of 400uM for compound 1, while the authors calculation gave a Kd of 350uM in Table 2.

Any mistake in my calculation?