The maximum affinity of methyls

Fragments are frequently evaluated in terms of the number of non-hydrogen atoms, and the lightest element commonly found in drugs is carbon. In a sense, then, a methyl group is the smallest possible fragment. Indeed, molecular modelers often use methyl groups as probes to sample a protein surface, and medicinal chemists fantasize about the “magic methyl” that will give a huge pop in potency. But how much affinity can a methyl group really give you? William Jorgensen and colleagues at Yale attempt to answer this question in a recent issue of J. Med. Chem.

The team analyzed every single paper published in J. Med. Chem. and Bioorg. Med. Chem. Lett. between 2006 and 2011 (ah, the joys of being a grad. student!) This produced a data set of 2145 examples in which researchers had replaced a hydrogen atom with a methyl group and reported dissociation or inhibition constants for both ligands; more than 100 different proteins were represented. Jorgensen and colleagues then plotted the change in free energy (∆∆G) for the hydrogen to methyl replacement.

The result was a roughly Gaussian distribution with a median of 0.0 kcal/mol. In other words, on average, adding a methyl neither improved nor decreased affinity. However, with a standard deviation of 1.0 kcal/mol, it was fairly common to get a 5-fold boost in affinity. But the frequency dropped off quickly from there: a 10-fold improvement occurred about 8% of the time, and the prized 100-fold boosts happened only 0.4% of the time – often enough to gain a persistent foothold in the imagination, but certainly not something you’d want to count on. To yield a 100-fold improvement, each methyl has a group ligand efficiency of 2.7 kcal/mol/atom, which is more than the Kuntz limit!

But higher boosts are possible: in a handful of cases the gain in potency was more than 180-fold, and much of the paper focuses on four of these truly magic methyls. A combination of crystallography and high-level molecular modeling revealed that ideal hydrophobic interactions were responsible for some of the affinity, but in all cases the methyl group also preorganized the conformation of the ligand to optimize interactions with the protein. The authors conclude:

It appears that to reach the 10-fold level, placement of a methyl group in a hydrophobic environment may be adequate; however, to go beyond that, the methyl group also generally needs to induce a propitious conformational change. This is typically achieved by ortho methylation in biaryl systems or by branching at an atom attached to a ring.

Interestingly, these conformational changes make the molecules less flat – more support for including three-dimensional fragments in your collection.

Experiences in fragment-based drug discovery

This is the title of a new review published in Trends in Pharmacological Sciences by Christopher Murray, Marcel Verdonk, and David Rees of Astex Pharmaceuticals. Although there is certainly no shortage of reviews on fragment-based lead discovery (a situation to which I have admittedly contributed), this one is notable both for its clarity and for being able to draw upon a deep wealth of institutional knowledge.

The review starts by discussing three notable case studies: Plexxikon’s discovery of the mutant B-Raf inhibitor vemurafenib, Astex’s Hsp90 program, and Merck’s BACE program.

Next, the authors describe some key concepts and challenges of FBLD.

Concept 1: Inappropriate physical properties are a major cause of attrition for small-molecule drugs

This should not come as a surprise to readers of this blog; the discovery of compounds with superior properties is one of the key selling points for FBLD. In support, the researchers compare 39 leads against 20 targets from Astex’s fragment-based programs with 335 published HTS-derived leads and 592 oral drugs. The FBLD-derived leads are on average 62 Da smaller and 1 log unit less lipophilic than are the HTS leads, and are much more similar to the oral drugs.

Concept 2: Although weak in potency, fragments actually form high-quality interactions

The position and the orientation of fragments tend to be conserved during the course of optimization (though see here for a notable exception). Of the 39 internal fragment-to-lead programs, roughly 80% of the atoms in the original fragment (which averaged 13 atoms total) were retained in the lead. Moreover, the mean shift in position as judged crystallographically was only 0.79 Å.

Concept 3: LE can be used to judge the relative optimisability of differently sized molecules

I like to think of fragments as ants: small and weak when considered from a human perspective, but impressively strong when considered for their size. Ligand efficiency and its many permutations are tools to assess molecules in a size-appropriate manner.

Concept 4: Relatively small libraries of fragments are required to sample chemical space

There is plenty of theory to support this (see for example here and here). The authors note that a library of 1000 compounds with 12 or fewer heavy atoms would sample ~0.001% of possible molecules with MW < 170 Da, while 1000 compounds with 25 or fewer heavy atoms would sample only 10^-14 percent of the possible larger molecules. But while theory is fine, the real proof is in the number of molecules that have entered the clinic that can trace their origins to small fragment libraries.

Of course, FBLD does have challenges.

Challenge 1: Specialized methods are needed to detect fragment binding

You don’t hunt ants with an elephant gun, and you’ll have a hard time finding fragments using standard procedures. The need for specialized methods was once a major impediment to FBLD, but happily today there are many options, and using two or more of these in combination is the best strategy.

Challenge 2: Efficient optimisation of fragment hits is required

In other words: you’ve found a fragment, now what? Structural biology is extremely helpful to figure out how the fragment binds and suggest what to do next, especially since proteins can be surprisingly flexible: of crystal structures from 25 fragment screens at Astex, 12 proteins showed movement of at least 5.0 Å upon fragment binding.

Of course, it takes more than a crystal structure to advance a fragment, and the challenges can be institutional as much as scientific. But given the proven success of the technique, these are challenges worth facing.

Finally, it’s worth checking out the entire issue of Trends in Pharmacological Sciences, which is devoted to structure-based drug design. There are some nice papers by Zhaoning Zhu on BACE, Tom Blundell and colleagues on protein-protein interactions, Stephen Wasserman and colleagues on high-throughput crystallography, and lots more.

Cracking the ROCK

In this paper, Li et al.  report the first use of FBDD on Rho Kinase (ROCK).  ROCK 1 and 2 are involved in several pathological diseases.  These researchers (not in pharma) started their efforts aiming to target the hinge region of the ROCKs with 7 pyridyl containing compounds.  

These compounds were tested at 400 uM .  Cpd 3 (nicotinamide) had IC50 of 76 and 56 against ROCK 1 and ROCK2, respectively.  The Ligand Efficiencies (LE = DeltaG/HA) were 0.37 and 0.39.  The authors note that they used LE as a general guide during the optimization process.  While not uncommon, it is nice to see more and more people explicitly stating this.  Cpd 4 and 5 had reduced activity compared to Cpd 3 which suggested to them other potential hinge binders. 

Cpd 8 (5-aminoindazole) had good potency and great LE against ROCK 1 and 2 (181 uM and 120 uM, 0.51 and 0.53).  This led to SAR to explore the linkers and binding site.

 Cpd 10 had good potency, 59 and 36 uM respectively.  Cpd 12 was previously reported as a ROCK 2 inhibitor at 260nM.  The authors state that the discrepancy is due to differences in assay conditions.  However, no two-carbon spacers had been reported, so they decided to pursue this avenue.  The ethylene spacer (Cpd 18) proved optimal (0.65 and 0.67 uM, 0.40LE).  However, modeling suggested that it was not fully utilizing all possible hydrogen bonds. 

In order to exploit all possible hydrogen bonds, they used Cpd 20 as an alternate hinge binder.  As they started to explore this chemistry (Cpds 22-27), the pyridines were published in a patent.

Luckily, Cpd 22 (with the new hinge binder) had similar activity to Cpd 18 (1.15 and 0.26 uM for Rock 1 and 2).  22 has a longer hinge binder, but shorter spacer than 18.  22 loses some of the hydrogen bonds seen in the modeling with 18 (N.B. this is NOT X-ray data), but gains of new ones are made.  So it is a wash on net hydrogen bonding, but freer IP space.  Cpd 18 was crystallized and as expected bound in the ATP site.

They then probed the chirality of the molecules, Cpd 24 (S configuration) was 75x more potent than Cpd 23 (R) for ROCK2, but only 5x for ROCK1.  Fascinatingly, adding one more carbon to the spacer (Cpds 26 and 27) reverses this preference.  

They then tested two pairs of compounds in cells (11/18 and 23/24).  They showed activity in the cells that correlated with the in vitro data.  

This is good example of academic drug design, but is raises some questions: 1. is selectivity for ROCK1 and ROCK2 important, 2. are these compounds in a clear IP space, 3.  Why weren't the other compounds (like 23/24) crystallized, 4. What is the basis for the chiral preferences with a linker length dependence?

[Ed: sorry for funky font at the end...don't know why that is like that.]

Halogenated fragments stabilize mutant p53

Practical Fragments recently discussed using fluorinated fragments for ¹⁹F NMR, but there are other halogens out there – are these useful for constructing fragment libraries? SGX Pharmaceuticals had a collection of fragments enriched with bromine atoms, the thought being that this atom would facilitate crystallography. Halogens can also make productive interactions with proteins, including so-called “halogen bonds” to backbone carbonyl atoms or pi-systems. With this in mind, Andreas Joerger at Cambridge University and Frank Boeckler at Eberhard-Karls University and their colleagues have assembled and screened a “halogen-enriched fragment library.” Their results are reported in a recent issue of J. Am. Chem. Soc.

The library consists of 79 non-reactive, soluble aromatic compounds containing bromine or iodine. Because these elements are so large, the researchers used a modified rule of 3 – instead of a molecular weight limit of 300, they limited the fragments to no more than 22 heavy atoms (see also our recent post on this topic here). They then screened this library against the Y220C mutant form of p53, which contains a surface crevice that destabilizes the protein and contributes to cancer cell survival. Thermal shift assays were used as the primary screen, with hits being confirmed by 2D NMR and ITC. This resulted in the discovery of compound 3, which crystallography confirmed was making a halogen bond to a backbone carbonyl.

Modifying the amine substituent improved potency modestly, and building off the phenyl ring towards a nearby pocket improved the potency further, albeit at a cost in ligand efficiency. Still, this compound (PhiKan5196) does represent the most potent Y220C binder reported, and represents an order of magnitude improvement over previous work. Moreover, the molecule induces apoptosis in p53 Y220C containing human cancer cell lines but not in matched wild-type p53 cell lines. (Unfortunately the compound also appears to be generally cytotoxic.)

This library is an interesting approach in part because it is somewhat heretical: for various reasons most library designers exclude molecules containing bromine or, especially, iodine. That said, the thyroid hormones do contain iodine aplenty, and MEK kinase seems to have a predilection for bromine or iodine as well. What do you think? Are halogenated fragments a useful tool for certain targets, or an unproductive diversion?

Poll: how many atoms are too many?

The recent rant about average molecular weight (AMW) leads to the question of how large a molecule can be and still be called a fragment. The Rule of 3 sets an upper limit of 300 Da, but perhaps we should think instead in terms of number of heavy atoms. If we take Andrew Hopkin’s calculation that the mean heavy atom adds 13.286 Da to a molecule, this would set an upper limit of 22-23 heavy atoms, but is that already super-sized?

Now is your chance to weigh in – please vote (on the right of the page) on the largest fragments you would put into your library, and feel free to comment too.

Why NOT AMW

I have been thinking a lot lately about library design, especially after the roundtable at breakfast in SD. I found a rant from an old friend/colleague, from the very early days of FBDD for me (2002 or 2003).   With his permission, but no citation for obvious reasons, I am reprinting it here.  I find this particularly interesting as AMW is still being used as recently as last year in papers describing fragment library design. 

The Average Molecular Weight [Ed:AMW] is currently an accepted orthodoxy within the Medicinal Chemistry community, a role reinforced by the recent popularity of things like the Rule of Five. To be sure, molecules with large molecular weights are not typically observed to be successful drugs.

In all the rest of this note, I’d like to focus on the common scenario of selecting molecules for purchase or testing. I’ve often seen people apply AMW cutoffs or scalings to these processes. I’d like to show why this may be sub-optimal.

First, I contend that the number of heavy atoms may be a much better proxy for “size” than AMW. Certainly, if you want to discriminate against heavier elements like Phosphorus, Sulphur, Chlorine, Bromine and Iodine, then by all means, use AMW. But with an AMW contribution of 126, a molecule with a single Iodine atom would be considered heaver/less desirable than the same molecule with a C6H9N2O substituent! Again, if you have good reasons for wanting to suppress these elements, AMW is a (very) good way of doing that.

The case of the third row and beyond elements is fairly straightforward, but what happens within the organic group Carbon, Nitrogen and Oxygen. Does AMW vs natoms mean anything in there?

Imagine we are selecting molecules for an assay and imposed an AMW cutoff of 180 – admittedly very low. We would then ignore Aspirin, with an AMW of 180.66, and instead we would test the “lighter” .

But it gets worse. Imagine if the Lilly chemists looking for antidepressants had imposed an AMW cutoff of 308. That would have excluded Prozac, with an AMW of 309,

and instead they would have tested the “more desirable lighter molecule”with an AMW of just 306.

Again, an AMW cutoff or bias could see us miss Zyprexa, AMW 312.4 and instead use the “lighter” molecule AMW 308.

Or the even more lighter still Oxygen variant! Now, of course, no medicinal chemist faced with the choice between these molecules would choose the second one, that’s obvious. Logp estimates would probably eliminate molecules that are all or mostly carbon atoms.

But what the examples above show is that in any automated system that is basing decisions on AMW, there will be a systematic, albeit small, bias towards Carbon atoms and against Nitrogen and Oxygen atoms. Why? A single CH2 group contributes 14 to AMW, whereas an NH contributes 15, and a two connected Oxygen atom 16. As terminal groups, CH3 is 15, NH2 is 16 and OH is 17. A Pyridine Nitrogen contributes 14, but an aromatic Carbon contributes 13, t-butyl groups preferred over CF3. Carbon wins the low AMW contest every time.

Now, how significant is this. Probably very small in most practical applications. I’d say that if you are setting up some kind of automated process, and you have equivalent access to AMW and the number of heavy atoms, use the number of heavy atoms in order to eliminate any small, pro-Carbon bias.

Conclusion

We see that automated procedures using AMW instead of natoms, will not only systematically suppress elements like P, S, Cl, Br and Iodine, but may also work to drive out N and O atoms as well!