From substrates to fragments – or not

Recently we highlighted a paper in which enzyme substrates were deconstructed into component fragments and tested against an enzyme with unknown specificity. In a new paper in J. Am. Chem. Soc. a collaboration led by Karen Allen (Boston University), Frank Raushel (Texas A&M), and Brian Shoichet (UCSF) has performed a similar experiment to ask whether fragments could be used to identify substrates.

The researchers chose six enzymes from three different classes and collected various-sized fragments based on known substrates. These were then tested in functional assays to see whether they could be substrates or inhibitors. Stunningly, in most cases the fragments showed no activity against the enzymes; when activity was detectable, it was usually at least 100,000-fold lower than the natural substrate. Even subtle tweaks, such as removing a hydroxyl group, were enough to mess things up, as illustrated for adenosine deaminase (compare compounds 1 and 4). Breaking the substrate in two was sometimes better: compound 8 was turned over slowly by the enzyme, though its complementary fragment 9 had no effect on activity – positive or negative – when added to the assay along with compound 8 or the natural substrate.

Of course, functional assays are less sensitive than biophysical assays, but in the one case where the researchers tried soaking fragments into crystals of the enzyme they found that the fragments bound in a different manner than the substrate – echoing previous work deconstructing synthetic inhibitors of protein-protein interactions.

As the authors note, the remarkably sharp structure-activity-relationships (SAR) observed here could reflect a fact of nature: most enzymes need to be highly selective for their substrates to avoid mucking up cellular metabolism.

Moreover, the notion that two fragments, when properly linked together, can bind more tightly than the sum of their individual binding energies has been a primary motivator behind fragment-based lead discovery for more than 30 years. In a sense, this paper illustrates this principle in reverse. Indeed, it is possible for the energy gained by linking two fragments to exceed the binding energy of an individual fragment.

This is a nice study from which we can draw two lessons, one pessimistic, the other optimistic. On the down side, we are unlikely to be able to use fragments to predict the natural substrates of uncharacterized enzymes, at least on a general basis. As noted previously, this is not surprising: the concept of molecular complexity predicts that fragments should be fairly promiscuous, and we’ve seen time and again that fragment selectivity is not necessarily maintained during optimization.

On the positive side, this study beautifully illustrates that it is possible to achieve massive enhancements in affinity with relatively small changes. Beyond just the magic methyl effect, we’ve got the magic hydroxyl effect, the magic thiophene effect – heck – the magic fragment effect. Of course, these are retrospective analyses, and it’s easier to break things than make them. That said, folks at Astex demonstrated that it is possible to improve the affinity of a millimolar fragment a million-fold by adding just six atoms. Perhaps such opportunities are more general than we have previously dared to dream.

Enthalpy arrays revisited: PDE10A

Two years ago we highlighted enthalpy arrays: very tiny temperature sensors that measure the heat generated during the course of an enzymatic reaction. Molecules that compete with a substrate will alter the kinetics of the reaction and can thus be identified as inhibitors. In the original paper a number of fragment hits were identified against the phosphodiesterase PDE4A, but unfortunately none of these could be structurally characterized. In a new paper in J. Biomol. Screen. Michael Recht and colleagues at Palo Alto Research Center have teamed up with Vicki Nienaber and colleagues at Zenobia to apply enthalpy arrays to the neurological target PDE10A, and this time they were able to obtain numerous crystal structures.

The researchers started by confirming that literature reference compounds behaved as expected. Next, they screened all 16 of the PDE4A hits against PDE10A, including several that were quite weak against PDE4A itself. All of these were active in the enthalpy array assays, with Ki values ranging from 94 to 1400 μM and good ligand efficiencies. In fact, most of the fragments were more potent against PDE10A than the phosphodiesterase against which they were original screened – which perhaps touches on the question of fragment selectivity.

The researchers also screened an additional 85 fragments at a concentration of 2 mM, leading to 8 more hits. All 24 of the hits were then soaked into crystals of PDE10A, yielding 16 crystal structures of bound fragments – a respectable 67% success rate. Interestingly, fragments that produced structures were more potent (average K_I = 590 µM) than those that didn’t (average K_I = 1000 µM), and this difference was statistically significant.

All of the fragments bound at the active site, and fragment growing was used to improve the affinity of two of the fragments. This led to low or sub-micromolar compounds, albeit with a loss in ligand efficiency. These more potent compounds were also selective for PDE10A over PDE4A, though solubility limits precluded testing at very high concentrations.

The paper frankly discusses some of the limits of using enthalpy arrays. For example, since the fragment should be present at a higher concentration than enzyme, very tight binders would require unfeasibly low enzyme concentrations. This limits the practical range of the technique to inhibitors with K_Is ranging from ~500 nM to 2 mM. Also, as Morgen G observed in a comment to the last post, this is more of a biochemical assay (monitoring the heat of an enzymatic reaction) rather than what most people think of when you say the word calorimetry (monitoring the heat of binding, as in the case of isothermal titration calorimetry). Still, enthalpy arrays seem pretty cool; hopefully folks will warm to them.

Fragment Library Vendors (2014 Edition)

We have been updating a lot of lists recently.  One that I think has changed significantly, is the fragment library vendor list, last updated in 2010. As Dan said four years ago, FOB Chris Swain has done a great job of curating who is selling what.  Instead of duplicating efforts, I will just focus on what has changed and making some comments.  I am not going to list companies that have libraries you can access, only those that sell outright their libraries. 

What are the keys for purchasing a good library?  I think minimally, purity and aqueous solubility should be experimentally tested and guaranteed.  I note those vendors who specifically point this out, but one should not assume that those who don't also don't have this data.  Comments can be sent directly to me or made below, and I will update this list.

Some general thoughts: 

• There is no special sauce.  Every library is good and will work for you. It's the choice of screen and how you prosecute it after that makes the difference.
• You don't need no stinkin' IP. 

3DFrag Consortium (New 2014):  I think this ran its course.  While I think by and large it had great ideas I don't think it ever truly answered the question "Do 3D fragments work better (in some areas)?"

Analyticon (New 2014): This is another example of fragments from nature.  The utility of these types of libraries are still up for discussion. 

Asinex: "Inspired by Nature" is its tagline.  However, they do have focused libraries, for such targets as PPIs.  They have 3159 in this library.. 

Chembridge: The collection is now 7000+ compounds (was 5000).  The guarantee greater than 90% purity, but nothing about solubility. 

Chemdiv (New 2014): Their collection is almost 14,000 fragments.

Enamine: More than doubled in size, from 12,000 to more than 28,000. 

Iota: I think they were the first to regularly use nPMI in their compound assesment.  You can only look at their library under CDA. 

Key Organics: They have quite a few specialized fragment libraries: CNS, self-assembly, brominated, fluorinated, and chiral cyclic molecules, in addition to their main libraries.  They guarantee 95% purity, 1mM aqueous solubility, solubility up to 200 mM in DMSO, and with almost no overlap with the Maybridge collections (68 compounds).

Life Chemicals:They are now up to 47,500 fragment molecules (less than 300 MW), of course only 31,000 of these exist, the other 16,000 can be made upon request.  They have 3900 19F fragments.  In terms of those that have experimental solubility, there are 8200.  However, 75% are soluble at 1mM, and 60% at 5mM in PBS.   So, always read the fine print.  They are also the vendor for the Zen-Life library, another library based on nature. 

Maybridge: The grandfather of them all.  30,000 fragments in total.  The 2500 Diversity collection is guranteed soluble at 200 mM in DMSO and 1mM in PBS.  The NMR spectrum is available, but only in organic solvent. It is available in many formats, from powder to DMSO-d6 solution. 

Otava: 8800 fragments in general.  800 19F fragments.  And 575 chelating fragments, if you want a warhead and all the issues they bring with.

Prestwick: 2200 fragments.

Timtec: No number available, but also can be shipped in DMSO solution. 

Vitas-M: The least helpful website out there.  It's Voldemort Rule compliant and available in multiple formats: mg, mcmol, sets, DMSO solution, dry film.

Zenobia:  Several different collections of very small fragments.

In defense of ligand efficiency – and poll!

Last year we highlighted a provocative article from Michael Shultz in which he took aim at the concept of ligand efficiency (LE). As we noted at the time, he raised some good points, and I am the first to argue that there is value in questioning widespread assumptions.

However, in addition to questioning the utility of LE, Shultz also questioned its mathematical validity. He repeated the attack earlier this year by asserting that ligand efficiency was a “mathematical impossibility.”

This is incorrect.

To set the record straight, Chris Murray (Astex), Andrew Hopkins (University of Dundee), György Keserü (Hungarian Academy of Sciences), Paul Leeson (GlaxoSmithKline), David Rees (Astex), Charles Reynolds (Gfree Bio), Nicola Richmond (GlaxoSmithKline) and I have written a response just published online in ACS Med. Chem. Lett. demonstrating that ligand efficiency is mathematically valid.

One of the criticisms of LE is that it is more sensitive to changes in small molecules (such as fragments) than in larger molecules. However, this is a property of any ratio, and we show that the same behavior applies to more familiar examples such as fuel efficiency: a few blocks of stop-and-go traffic has more of an effect on the overall fuel efficiency of a short trip than a long trip.

Of course, that’s not to say that ligand efficiency and other metrics are perfect or universally applicable; we discuss a number of situations where they may be more or less useful.

In this spirit, Practical Fragments is revisiting a poll from 2011 to see what metrics you use – please vote on the right-hand side of the page, and share your thoughts here. Note that you can vote for multiple metrics, and please check the last box (Polldaddy does not tally individual responses, so this box will track total number of voters to allow us to calculate percentage of respondents who use a given metric).

Keep the comments coming, and check back to see the poll results.

Biofragments: extracting signal from noise, and the limits of three-dimensionality

What does this protein do? Now that any genome can be sequenced, this question gets raised quite often. In many cases it is possible to give a rough answer based on protein sequence: this protein is a serine protease, that one is a protein tyrosine kinase, but figuring out the specific substrates can be more of a challenge. In a recent paper in ChemBioChem, Chris Abell and collaborators at the University of Cambridge and the University of Manchester attempt to answer this question with fragments.

The bacterium Mycobacterium tuberculosis (Mtb), which causes tuberculosis, has 20 cytochrome P450 proteins (CYPs), heme-containing enzymes that usually oxidize small molecules. Although some are essential for the pathogen, it is not clear what many of them do. The researchers used an approach called “biofragments” to try to pin down the substrate of CYP126.

The biofragments approach starts by selecting a collection of fragments based on known substrates. Of course, the specific substrates are not known, so in this case the researchers started with a set of several dozen natural (ie, non-synthetic) substrates of various other CYPs, both bacterial and eukaryotic. They then computationally screened the ZINC database of commercial molecules for fragments most similar to these substrates and purchased 63 of them. Perhaps not surprisingly given their similarity to natural products, these turned out to be more “three-dimensional” than conventional fragment libraries, as assessed both by the fraction of sp³ hybridized carbons and by principal moment-of-inertia.

Next, the researchers screened their fragments against CYP126 using three different NMR techniques (CPMG, STD, and WaterLOGSY). Since they were primarily interested in hits that bind at the active site, they also used a displacement assay in which the synthetic heme-binding drug ketoconazole was competed against fragments. This exercise yielded 9 hits – a relatively high 14% hit rate.

Strikingly, all of the hits are aromatic, and 7 of them could reasonably be described as planar. In other words, even though the biofragment library was relatively 3-dimensional, the confirmed hits were some of the flattest in the library! The researchers interpreted this to mean that “CYP126 might preferentially recognize aromatic moieties within its catalytic site,” but there could be something more general going on – perhaps aromatics are simply less complex, and thus more promiscuous.

Examining the fragment hits more closely, the researchers found that one of them – a dichlorophenol – produced a spectrophotometric shift similar to that produced by substrates when bound to the enzyme. This led them to look for similar structures among proposed Mtb metabolites. Weirdly, pentachlorophenol came up as a possible hit, and a spectrophotometric shift assay reveals that this molecule does have relatively high affinity for CYP126. Whether this is a biologically relevant substrate for the enzyme remains to be seen.

This is an intriguing approach, but I do have reservations. First, in constructing fragment libraries based on natural products, it is essential to avoid anything too “funky”. The Abell lab is one of the top fragment groups out there, well aware of potential artifacts, and has a long history of studying CYPs, but researchers with less experience could easily populate a library with dubious compounds.

More fundamentally though, I wonder about the basic premise of biofragments. The whole point of fragments is that they have low molecular complexity and are thus likely to bind to many targets, so is it realistic to try to extract selectivity data from them? Indeed, as we’ve seen (here and here), fragment selectivity is not necessarily predictive of larger molecules.

That said, the approach is worth trying. Even if it doesn’t ultimately lead to new insights into proteins’ natural substrates, it could lead to new inhibitors.

Ninth Annual Fragment-based Drug Discovery Meeting, Part 2

The first major fragment event of 2014 drew around 500 people to San Diego last week. This is part of CHI’s three-day Drug Discovery Chemistry conference, and although the official FBDD track was only one of six, it is a testimony to the vitality of the field that fragments made appearances in most of the other sessions. With 17 talks in the FBDD track alone this post will not attempt to be comprehensive; Teddy has already shared some impressions here.

Jim Wells (UCSF) gave a magisterial keynote address that emphasized how useful fragments can be for tackling difficult targets such as protein-protein interactions (PPIs). In fact, many of the talks in the protein-protein interaction track relied on fragments. That’s not to say it’s easy. Rod Hubbard (University of York and Vernalis) emphasized that advancing fragments to leads against such targets can take a long time and often requires patience that strains the management of many organizations. Fragment hits against PPIs usually have lower ligand efficiencies (0.23-0.25 kcal/mol/HA if you’re lucky), and improving potency can be a bear. Rhian Holvey (University of Cambridge) presented a nice example of how she was able to find millimolar fragments that bind to the anti-mitotic target TPX2, potentially blocking its interaction with importin-alpha, but even structural information was not enough to get to potent inhibitors.

G-protein coupled receptors (GPCRs) were thought to be unsuitable for fragments until recently, but both Iwan de Esch (whose work has been profiled several times, including here and here) and Jan Steyaert (Vrije University) presented success stories. In fact, Jan has only been working with the Maybridge fragment library for a few months, but has found agonists, antagonists, and inverse agonists for several GPCRs.

Another example of a difficult target is lactate dehydrogenase A (LDHA). We’ve previously highlighted cases where fragment linking was used to get to nanomolar binders (here and here); Mark Elban (GlaxoSmithKline) presented an example of fragment growing and using information from a high-throughput screen (HTS) to get to nanomolar binders. Mark also discussed a particularly disturbing false positive: HTS had generated dozens of confirmed hits spanning 7 chemotypes, but upon closer inspection it turned out that all of them came from a single vendor, and that – unreported by the vendor – they were all oxalate salts. Oxalate is a low micromolar inhibitor of LDHA, and is invisible in proton NMR, so I’m sure this was not fun to track down.

Ben Davis (Vernalis) also presented great examples of false positives and false negatives, and how to avoid them. In particular, the WaterLOGSY NMR technique is great for weeding out aggregators when run in the absence of protein.

A common theme throughout the conference was the integration of fragments with other methods, such as HTS. Nick Skelton (Genentech) actually titled his presentation “Fragment vs. HTS hits: does it have to be a competition?” Kate Ashton (Amgen) discussed how using information from a fragment screen helped solve pharmacokinetic issues with an HTS-derived hit. And Steven Taylor (Boehringer Ingelheim) presented a similar example (also covered here) of using fragments to fix a more advanced lead. Steven noted that fragment-based methods are now fully integrated into the organization, which marks a significant change from Sandy Farmer’s presentation at this meeting four years ago.

The roundtables are great opportunities to swap ideas and get feedback; Teddy already mentioned the excellent roundtable he chaired, but I wanted to also give a shout-out to one organized by Derek Cole (Takeda) focused on "practical aspects of fragment screening." We recently discussed discussed fragments that destabilize proteins in thermal shift assays, and it turns out that folks from both the Broad Institute and Takeda have also crystallographically characterized such fragments. There was the sense that either stabilizers or destabilizers should be considered hits, though the latter were less likely to lead to crystal structures than the former.

Finally, on the subject of library design, Damian Young (Baylor College of Medicine) described using diversity-oriented synthesis (DOS) to generate more “three-dimensional” fragments. He is planning to build a library of roughly 3000 fragments which he hopes to make widely available to the community; these should help answer the question of whether the third dimension is really an advantage.

The importance of library design was also emphasized by Valerio Berdini (Astex); they are currently on their seventh generation library, about 40% of which is non-commercial, and half of whose members have been solved in one or more of 6000+ crystal structures. Relevant to the rule of three, Astex is moving to ever smaller fragments, with an average of 12.6 non-hydrogen atoms, ClogP = 0.6, and MW = 179. Indeed, despite assertions that PPIs may require larger fragments, Rod noted that at Vernalis the average fragments hits against PPIs are only slightly larger (MW = 202 vs 189 against all targets) and more lipophilic (ClogP 1.2 vs 0.8).

CHI has already announced that next year’s meeting will be held in San Diego from April 21-23. As it will be the ten year anniversary, they’re planning something big, so put it on your calendar now!

Drug Discovery Chemistry Conference Round Up, Pt 1

As many of you know, Dan and I were at the CHI Drug Discovery Conference.  Over the next few posts, we will be posting notes, thoughts, and some comments on what happened. First off, I live-tweeted the sessions I was at.  Day 1's highlights are here , Day 2 is here, Day 3 here.  Overall, I really liked the conference and thought the agenda was very high quality (one session was poor, read my tweets and guess which one).  CHI had Dan and me hopping: chairing sessions, moderating round tables, and co-teaching our FBDD course.  I was very lucky to moderate a breakfast roundtable  on Thursday morning on PPIs and Fragments (Small Solutions for Big Problems: Fragments and PPI).  

The table was a huge hit; there were more than 30 people spread over 5 or so tables.  I am pretty sure it was because Rod Hubbard and Dan were there.  I am going to try to capture the discussion, but I am pretty sure I am missing key points, so if people where there and remember things, add them in the comments, or email me and I will edit the post.  The discussion initiated with a reference to a comment that Rod made in his talk the day before: NMR is the preferred method for screening PPIs (over SPR).  He cited two main reasons: NMR is more sensitive to very weak binders and you can do QC on the protein in every samples.  With SPR, once the complex is put on the chip, you have no idea if it is still intact/amenable to screening.  Jan Steyaert asked a question: most of the PPIs we see targeted are stable complexes, while most of the PPIs in nature are transient, why?  This led to the question of whether people are working to stabilize complexes, rather than disrupt?  People agreed that you would need to have a kinetically resolved assay, like a TR-FRET.  I raised the point that you would need a enzymologist to do this, and most biologists these days are pharmacologists.  [As an aside, this is why we focus so strongly on IC50 without knowing really what it means.  For further discussion of this, go ask Pete Kenny about this.]  It was brought up that you could possibly due this by SPR if you  had a a very special "group"[my handwriting sucks under the best of cases, rushing while moderating makes it even worse].  I think the idea is you could go to lower temperatures and start observing kinetics of fragment binding (vs. the normal square sensorgram). 

Someone from the Broad Institute said that they use Thermal Shift for PPIs, no matter what.  However, most of the people don't trust TS, no matter what.  It is cheap and fast, but so full of artifacts and errors.  It's a paradox, a quick, cheap assay for PPIs that everyone uses and no one trusts.  Seems like a classic case of the herd mentality.  

The discussion then moved on to a key concept for fragment/PPIs: how do you follow up on 3D fragments?  Most people agree that fragments with higher 3D content are better for targeting PPIs.  However, I think, in contrast to 2D fragments, exactly how you prosecute hits in this target class is less straight forward.  I think an important distinction between 2D and 3D fragments in terms of follow up is that linking 3D fragments may actually be relevant and productive, in contrast to how most people view linking 2D fragments.  

As I said above, I am sure I am missing some points, so add them in the comments or email me.  Dan and I will be posting more round up notes over the next week or so, so stay tuned.

Fragments vs soluble epoxide hydrolase – hundreds of them!

Soluble epoxide hydrolase (sEH) is a potential target for cardiovascular and immune disorders. The enzyme catalyzes the hydrolysis of long chain, lipophilic epoxyeicosatrienoic acids. These bind in a largely hydrophobic “L-shaped” pocket, with the catalytic machinery at the point of the L. Although it is relatively easy to find potent inhibitors of this enzyme, these tend to be greasy, insoluble, and non-druglike. In a recent paper in Bioorg. Med. Chem. Yasushi Amano and colleagues at Astellas describe a fragment-based approach to find better leads.

The researchers started with a high-concentration (up to 2 mM) enzymatic inhibition assay of 4200 fragments, resulting in 307 hits with IC₅₀ values between 700 nM and 1.7 mM. All of these were taken into co-crystallography trials, yielding crystals for about half of them. The other fragments were soaked into apo-crystals of sEH (that is, crystals without bound ligand) to get as many structures as possible. All together 126 crystal structures of fragments bound to sEH were solved, which is all the more impressive considering that there are only three authors on the paper!

Most of the fragments (83) bound at the catalytic site (example shown in green), while 29 bound to one of the lipophilic branches of the L and 9 bound to the other (cyan and magenta). Five fragments bound at two different sites within the enzyme. The researchers discuss ten of the fragments in some detail.

Many of the fragments that bind at the catalytic site contain amides or ureas – moieties in known inhibitors – but some of them (such as compound 3 above) are more unusual. Also, despite the generally lipophilic nature of the binding pocket, many of the fragments – even those that bind in the hydrophobic branches of the L – make hydrogen bonds to the protein or to bound water molecules. This suggests that it should be possible to find potent inhibitors that are less hydrophobic than previously reported molecules. Fragment growing, linking, and merging approaches could all work, and indeed the researchers hint that results of these studies will be reported in future papers.

More importantly, this paper provides a great set of crystallographically validated fragments binding to distinct sites on a well-characterized protein. Helpfully, the researchers have deposited the coordinates of the 10 co-crystal structures discussed in the protein data bank. Moreover, all of the fragments are commercially available. This seems like an ideal model system for validating computational docking and scoring approaches as well as for better understanding the energetics of protein-ligand interactions. If I were an academician working in these areas, I’d jump in feet first!

What we do in life, echoes in eternity (or the life of the patent)

Next week is the Drug Discovery Chemistry conference where Dan and I will be co-teaching our award-winning FBDD short course (or at least our mom's think it is great). We look forward to seeing any/all of you next week.  Blogging may be light next week, but we promise to give an update of the going-ons at the conference soon after.  

Kinases are fun, and those of us who have worked in them have probably all worked on the same ones.  I always loved the MAP family.  Why would I have a favorite kinase family?  Because of the cascading MAP kinases, like the one in this paper, Mitogen-activated protein kinase kinase kinase kinase 4 (that's a lot of kinase!).  But, unlike a lot of other kinases, there is no good tool compound.  So, using SPR, they decided to generate one. This paper is not particular interesting in terms of what they did, but rather it raises interesting questions. While the approach they describe is not novel, it is nice to see the data supporting them. 

They screened their 2500 fragment library against immobilized protein at 100 uM (single point).  225 hits were found with Kd ranging from 10 to 2000 uM (LE =0.24 to 0.59) for a 9% hit rate.This paper is about progressing this oxazole fragment 1: 

Based upon its structure and the wealth of kinase structure knowledge extant, they surmised it would be ATP-competitive and a hinge binder.  Based upon a binding model, the attempted to prosecute this fragment by "close-in" analogs and looking for groups that would extend farther into the hydrophobic pocket, but with MW less than 350 Da and clogP less than 3.5.  Exploring bi-aryl space resulted in 8:  

This compound had an activity of 143 nM and it was at this point that they decided to switch to the biochemical assay as their primary assay.  In the end, using X-ray focusing on LLE, they ended up deliveringa low molecular weight compound with favorable in vivo PK.  It also demonstrated a pathway functional response. 

This raises an excellent point, something I get asked frequently.  When do you switch from a biophysical assay to a biochemical one?  This maybe arguing semantics, but I think as more and more companies enter this arena these are exactly the things we need to discuss.  I think the switch happens when you feel comfortable, there is no hard and fast rule.  There is a difference in the SPR Kd and biochemical IC50 by more than 10x.  It is very important to note that they relied heavily on LE (-RTlnKd/HA) and LLE (pKd-cLogP), or pIC50 for biochemical assays.  But, it also raises the issue of correlation between SPR Kd and IC50.  I raise these socratically, and as maybe as a topics for discussion next week (or in July and September).   

Can selectivity of fragments be maintained?

In 2011 we highlighted an analysis of kinase inhibitors that demonstrated that non-selective fragments could produce selective leads, and vice versa. However, that study was based on hundreds of compounds not necessarily chosen from the same projects. Are the results the same within individual fragment-to-lead programs? This is the question that Ian Collins and colleagues at the Institute of Cancer Research address in a recent paper in MedChemComm.

The researchers examined three fragment-to-lead efforts: two targeted the kinase PKB and the other targeted the kinase CHK1. In all three cases they started with fragments and used structure-based design and fragment growing to obtain low nanomolar inhibitors. Importantly, they also obtained crystal structures of key compounds along the way, demonstrating that the initial fragment – a hinge-binding element – maintained its position and orientation throughout the process.

Each fragment, lead compound, and intermediate molecule was tested for selectivity in a panel of 91 kinases using a microfluidic mobility-shift peptide phosphorylation assay. The concentration of ATP in each assay was at the K_M,ATP, and each compound was tested at 10-fold above its IC₅₀ for the target kinase (so for example fragment 1 below was screened at 1000 μM, and fragment 5 was screened at 8 µM). For each compound a selectivity score was calculated based on the number of kinases inhibited at a certain threshold. For example, if S_(30%) = 1, this would mean that all of the kinases were inhibited by at least 30% at the concentration tested, whereas if S_(30%) = 0.03 this would mean that only 3 kinases (3/91= 0.03) were inhibited.

It's worth noting that selectivity is tough to define specifically. Although the selectivity score makes sense intuitively – each compound is tested at a concentration relevant to the intended target – I am concerned that it will make potent compounds appear more selective than they really are. Indeed, a plot of S_(30%) versus -log[concentration tested] is fairly linear. (Compounds discussed below are labeled by number on the plot.)

Accepting this definition of selectivity, though, it appears that nonselective fragments, such as 7-azaindole (fragment 1) could be progressed to nonselective leads such as compound 4, which was an early milepost en route to AZD5363, currently in Phase 2 clinical trials.

Fragment 1 was also modified to slightly less promiscuous fragment 5. A slight tweak to this molecule produced selective fragment 6, which was then optimized to the selective compound 8 (closer to AZD5363). In the case of fragment 6, even though the structural change was minor (removal of a single methylene) this was enough to make a specific interaction with a residue in PKB not found in other kinases. The CHK1 story is similar in taking a nonselective fragment to a selective lead.

So what’s the conclusion? The authors suggest that:

Broad kinase selectivity screens of fragments could be predictive of the lead, provided strategies to conserve the profile are followed in the elaboration, avoiding introducing new interactions with target-specific residues. Conversely, the initial fragment selectivity patterns are unlikely to reflect those of developed leads if the fragment does not already encode the anticipated target-specific interactions.

In other words, it depends. This is not meant as a criticism: I think this conclusion is about as decisive as possible when generalizing about fragment-to-lead strategies. At the very least, the work suggests that effort spent optimizing a fragment before growing or linking could be worthwhile. And even a promiscuous fragment may be only one atom away from something quite specific.

Covalent, destabilizing fragments against TB target BioA

Differential scanning fluorimetry (DSF) is a hit-finding technique in which a protein is incubated with a small molecule and heated until the protein reaches its “melting temperature” and unfolds. If a small molecule binds, in theory it should stabilize the protein towards thermal denaturation and thus raise the melting temperature, giving a positive thermal shift. However, most folks who have performed these types of experiments have also found (and usually ignored) molecules that lower the melting temperature of the protein. In a new paper in ChemBioChem, Barry Finzel and coworkers at the University of Minnesota follow up on one of these with very interesting results.

The researchers were interested in the enzyme 7,8-diaminopelargonic acid synthase (BioA) from Mycobacterium tuberculosis, the organism that causes its eponymous disease. This enzyme – which is not found in mammals – is involved in the synthesis of the essential cofactor biotin. DSF was used to screen 1000 compounds from the Maybridge Ro3 Diversity Fragment Library at 5 mM concentration, resulting in 21 hits which changed the denaturation temperature (T_m) by more than 2 °C. A dozen of these decreased the T_m, but although all of these were taken into crystallography trials, only compound 1 yielded a structure. STD NMR was also used to confirm that the compound binds to BioA in solution.

Next, the researchers used the classic “SAR-by-catalog” approach and purchased analogs of compound 1. Compound 2 turned out to be particularly interesting: it decreased the T_m by a whopping 18 °C! Weirder still, when soaked into crystals of BioA, they turned from yellow to red.

BioA is a transaminase: it takes a nitrogen from one molecule (called SAM) and transfers it to another molecule (called KAPA). En route to its final destination, the nitrogen is transferred to a co-factor, pyridoxal phosphate (PLP), which contains an aldehyde. Compound 2 contains a hydrazine, which is known to react with aldehydes, and in fact a co-crystal structure of compound 2 bound to BioA shows that this is exactly what happens. Interestingly, compound 2 binds in a somewhat different manner than compound 1 despite their similar chemical structures.

Compound 2 turns out to be a reversible inhibitor of BioA, and the researchers were able to demonstrate that it is a moderately potent and competitive inhibitor with respect to SAM and a less potent uncompetitive inhibitor with respect to KAPA. This is exactly what you would expect, since it competes with SAM for binding to PLP but does not compete with KAPA.

Now you may think that hydrazines aren’t exactly drug-like, but it turns out that a commonly used drug against tuberculosis is isoniazid, which contains an analogous acyl hydrazide. The researchers found that isoniazid also decreases the T_m of BioA, though less dramatically than compound 2. Though isoniazid works through an entirely different mechanism, the researchers were able to obtain a co-crystal structure of this binding to PLP in BioA (magenta; PLP is on the left, and protein is not shown), showing that it binds differently than either compound 1 (green) or 2 (cyan). Nonetheless, it did not show any inhibition of the enzyme, demonstrating that covalent binding alone is not sufficient for disrupting enzymatic activity.

This is a very nice paper, and it will be fascinating to try to understand how the fragments so effectively destabilize the protein despite binding tightly, and how this translates into inhibition. The researchers suggest that finding ligands that destabilize proteins could be generally useful for turning off proteins. Are there other well-characterized examples out there?

It's A Start

As the readers of this blog know, I tend to be harsh on academic "Drug discovery" papers.  Sometimes, there is a really worthwhile academic paper, but by and large I find that they tend to publish things that are barely "drug discovery" and more the For Dummies...of what they think drug discovery is.  Which way will I swing on this paper from researchers Down Under?  

The bacterial Sliding Clamp, aka polymerase 3beta, is a key player in bacterial replication and is an "emerging" target. It interacts with other proteins via LM (Linear Motifs): 4-10 amino acid disordered regions.  This is typically a weak interaction (1-100 uM). These LM exist at termini, but sometimes in loops.  A consensus sequence for the LM that interacts with the Sliding Clamp has been identified: QLx1Lx2F/L (S/D preferred at x1; x2 may be absent).  Two classes of compounds have been identified previously but with >10 uM affinity and no -cidal activity. 

So, these authors went after this target using X-ray as the primary screen.  The Zenobia Fragment Library was used (352 molecules) to soak into crystal in pools of 4 fragments.  Four fragments (below) were found to bind to Subsite I on Chain A.  However, no changes in the main chain density were observed.

They also found several other fragments with weak density, and several that were deemed crystallographic artifacts.  None of these compounds showed significant activity below 1 mM in their competition assay.  So, the story then continues that they "sought to improve binding affinity by identifying fragments that could more completely occupy" the binding site.  

[An aside:  To me, this brings up an important point about the choice of fragment collection.  Fragments that are designed for X-ray soaking tend to be small (10-12 HAC).  Just from a theoretical standpoint, those fragment would have to have an affinity in the 250 uM range (LEAN 0.3).   This was covered in a poll and most most people are happy going < 10 HAC.  My question is how often is a very small fragment found as an active?]

To do this, they noted that the fluoro-phenyl group in 1 was previously reported, leading to investigations with compound 5. It was found to fully occupy the binding site.

They searched ZINC for compounds similar to 1-5.  Their initial purchases failed to find any compounds with activity < 1mM.  Eventually, they landed on the hypothesis that chlorocarbazoles were "promising", leading to compound 6.  At this point, I think Dan's head exploded, Scanners-style.  Yes, that is an epoxide.  The co-crystal structure showed that it was binding in the active, albeit with weak electron density.  Their SAR, wisely, did not include the N-alkyl epoxide. 

Both 7 and 8 show good LE and LLEAT.  Only the R enantiomer of 8 caused movement in the main chain.  It also was the most potent in the replication inhibition assay (64 uM).  It was also the most potent in terms of -cidal activity against both Gram positive and negative microbes. 

So, is this a good or bad paper?  I would say it is a start, but if I had been a reviewer I would have made them change the title "Discovery of Lead Compounds Targeting the Bacterial Sliding Clamp
Using a Fragment-Based Approach" to "Discovery of ACTIVE Compounds Targeting the Bacterial Sliding Clamp Using a Fragment-Based Approach".

Funky fragments

Natural products have led to many approved drugs, and there is an increasing appreciation that Nature often knows best. Indeed, several published fragment libraries incorporate natural products or natural product-like molecules (see for example here, here, and here). With all this attention, it was inevitable that commercial fragment suppliers would spot this market need.

SerpentesOleum, Inc. has just launched a library they call FUNK: Fragments Uncovered in Natural Kompounds. This set consists of several hundred natural products and derived fragments carefully selected to maximize hit rates. For example:

The company has screened their library against targets such as PTP1B and falcipain-1 and obtained remarkably high hit rates in functional assays. In fact, SerpentesOleum is so confident that they’re offering a money-back guarantee if you don’t obtain at least one active against your target, no matter what it is. Looking at the structures of their compounds, I have no reason to doubt their claim.

Who's Doing FBLD, 2014 Version

It's been a while since we updated this list.  The first list had 24 companies, the second 44.  From 2011 to now, what's changed?

I have divided the list into (primarily) providers of services and pharmaceutical companies. This does not mean that that providers don't do their own discovery, or work in a mixed-model.  I have annotated those companies which were not previously on the list.  Please note that this does not necessarily mean they were not doing FBDD previously, they just were not listed.If we missed someone, let us know and we will update the post.

Providers:
Ancorex  (New 2014)
Beactica
Biodesy  (New 2014)
Biofocus(Galapagos)  Acquired by Charles River Laboratories
Biosensor Tools
BioSolveIT
Chemical Computing Group   (New 2014)
Crelux
CrystaX Pharmaceuticals  Acquired by Oryzon Genomics
Domainex   (New 2014)
Emerald BioStructures
Exscientia   (New 2014)
Evotec
Graffinity  Acquired by NovAliX 
iNovacia  Acquired by Kancera
Infarmatik
Intellisyn (New 2014)
IOTA Pharmaceuticals
Kinetic Discovery
MEDIT
Molsoft (New 2014)
Nanotemper (New 2014)
NMR Research  (New 2014)
NovAliX
Pharma Diagnostics
Proteros 
Pyxis Discovery  (Are they still an ongoing concern?)
Red Glead  (New 2014)
Saromics (New 2014)
Schrodinger
Selcia
SensiQ  (New 2014)
Structure Based Design
Viva Biotech  (New 2014)
Zenobia Therapeutics
ZoBio

Companies: 
Abbvie             split from Abbott
Amgen             (New 2014)
Ansaris (previously Locus)
Ariad               (New 2014)
AstraZeneca
Astex              Acquired by Otsuka(2013)
BioLeap
Boehringer Ingelheim
Bristol Myers Squibb
Carmot Therapeutics
Constellation Pharma
Crown Biosciences 
Dart Neuroscience
Eli Lilly
Genentech (Roche)
Genzyme Acquired by Sanofi-Aventis
GlaxoSmithKline
Heptares
Johnson & Johnson 
Merck
Nerviano Medical Sciences
Novartis
Pfizer
Plexxikon    Acquired by Daiichi Sankyo (2011)
Polyphor
Roche
Sprint Bioscience
Takeda California
UCB   (New 2014)
Vernalis
Vertex

**UPDATE** 27Mar2014: added Domainex, UCB
**UPDATE2** 28May2014: added MolSoft, Ancorex

Fragments vs MCL-1, again and again

Last year we highlighted a paper from Stephen Fesik’s group at Vanderbilt in which he used SAR by NMR and fragment merging to identify nanomolar inhibitors of the protein MCL-1, an anti-cancer target that had previously been thought to be impervious to small molecules. In a recent paper in Bioorg. Med. Chem. Lett., Andrew Petros, Chaohong Sun, and other former colleagues of Fesik at AbbVie describe two additional series of inhibitors.

The researchers started with an NMR screen using MCL-1 in which the methyl groups of isoleucine, leucine, valine, and methionine were ¹³C-labeled. Screening this against a library of 17,000 fragments in pools of 30(!) gave dozens of hits, some of which inhibited in a biochemical assay (for aficionados, they assessed binding to the BH3 domain of Noxa using fluorescence polarization as a readout).

Fragment 1 turned out to be fairly potent, though it is super-sized and violates the rule of three. The researchers were unable to get co-crystal structures of any of their fragments bound to MCL-1, but they were able to use NOE-based NMR experiments to develop a model of how fragment 1 might bind. This led them to synthesize a number of analogs such as compound 17, for which they were able to obtain a co-crystal structure with the protein, ultimately leading to the mid-nanomolar compound 24.

Fragment 2 was much less potent than the other fragment but had a considerably higher ligand efficiency. In this case simple modeling suggested growing away from the acidic portion of the molecule, leading to compound 36 (which was characterized crystallographically bound to MCL-1) and the more potent compound 44.

Overlaying the co-crystal structures of compounds 17 (blue) and 36 (red) reveals that they both bind in the same region, where Fesik’s compound 53 (green) also binds. All three molecules place a carboxylic acid in a similar position, but the two more potent molecules thrust a hydrophobic moiety deep into a pocket of the protein. It is tempting to speculate that compound 44, the more potent analog of compound 36, may also take advantage of this pocket.

Andrew Petros presented some of this work at FBLD 2012, so it is nice to see it in print. Though reasonably potent, it is worth keeping in mind that the molecules are also quite lipophilic. Perhaps it is significant that, like the Fesik paper, no cell-based data are presented. Collectively, though, these papers establish that MCL-1 is ligandable. Whether it will be druggable remains an important – and as yet unanswered – question.

PAINS propagation

PAINS, or pan-assay interference compounds, comprise a subject that has cropped up several times here (and here, and here, and here). Although not exclusive to fragments, I wanted to point out a thorough and insightful analysis by Jonathan Baell over at HTSPAINS in which he traces the lineage of a dubious series through paper after paper all the way back to 2001. The assays, models, and mechanistic theories all change, and the molecules keep getting uglier as they devolve from chalcones to bis-benzylidenepiperidones. It’s an entertaining and educational look at sloppy science. He ends with an important point:

People still don’t realize how easy it is to get a biological readout. The more subversive a compound, the more likely this is.

Indeed, even with decent looking molecules it can be difficult to figure out exactly what is going on; with PAINS you may as well start explaining things in terms of phlogiston and humorism.

I can’t help thinking of the late Efraim Racker’s admonishment: “don’t waste clean thinking on dirty enzymes.” Even when they are chemically pure, PAINS molecules are mechanistically dirty. The amount of effort wasted on them boggles the imagination, so keep them out of your libraries!

This is another way to do it.

The key to doing something right is to following the directions.  How closely you follow the directions, or don't follow, can be the difference between brilliance and just a good performance, e.g. cooking.  Sometimes, directions are meant as guidelines, like the Pirate Code or the Voldemort Rule.  Late last year, and blogged about here, I published a paper in Current Protocols on how to prosecute an STD screen.  A recent paper in PLOSOne, shows how someone else runs their screens, but with details on library construction, solubility testing, and more.  What makes this paper of interest is the level of detail that they provide.

Library Design: They assembled a diverse fragment library with the following rules: 110≤ molecular weight ≤350, clogP≤3, number of rotable bonds ≤3, number of hydrogen bond doners ≤3, number of hydrogen bond acceptors ≤3, total polar surface area ≤110, and logS_w (aqueous solubility) ≥ −4.5. 

I am little confused by the figure and what the text says.  In the text, they seem to have relaxed the MW cutoff, but the figure shows that anything not Voldemort Rule compliant is tossed.  They also preferred that the compound has at least one aromatic peak (for easier NMR detection).  They purchased 1008 from Chembridge, solubilized at 200 mM in DMSO-d6 (ease of NMR detection, again) and then tested the solubility at 1 mM in water.  I would have added some salt here, 50 mM, but that is a quibble.  For purity, they claim a low level of impurity (< 15%)!!!  To me, this is a whole lot of impurity.  But, as has been noted here, purity levels vary from library to library.

Solubility Testing:  They then made sure to experimentally test every fragment for solubility.  I can agree more emphatically with this approach.  Bravo!  They go into great detail, which I will not attempt to replicate here, but thanks to open access, they have included the scripts in the supplemental.  Acceptable compounds had > 0.1 mM aqueous solubility.  For me, this is too low, but to each their own.  They ended up with 893 total fragments (89% passed).  The real data I would like to see is how many fail if the cutoff is set at 0.5 mM or higher.  

Pooling: They then describe their pooling strategy.  I like open access articles for a lot of reasons, and tend to overlook small editorial problems (typos, grammar, etc.), but in this case, let me rant.  The authors state in the text that a random mixing of compounds would lead to severe overlap, exemplified in 3a.  To me, it does no such thing. 

Their approach is very similar to the Monte Carlo-based one that has previously been discussed on this blog.  Their final pools contain 10 fragments at 20 mM (I assume in 100 % DMSO-d6). 

Screening: They also acquired the 1H spectrum, STD (-0.7 ppm, > 1 ppm from any methyl), and WaterLOGSY spectrum of every pool for future reference.  This is a very clever approach as the STD should give no signal while the WaterLOGSY should give inverted peaks for all compounds in the pool (when interacting with a target they will be "right-side up").  Again, the figure may show that (I think if you blow up the figure the WaterLOGSY spectra does have peaks) but it is very difficult to see. 

Three of the 90 pools (3.3%) showed peaks in the aromatic region, most likely due to aggregation (they observed precipitation).  I would like to know if those compounds showed STD peaks also had those methyl groups within 1 ppm of the saturation frequency.  I would also like to know if they removed those compounds from the library, or just dealt with it.  For a paper with a great level of detail, it falls flat in this respect.  

Screening is performed at 10uM Target: 500uM ligand and the following parameters: acquisition time of 1 s, 32 dummy scans, and relaxation delay of 0.1 s, followed by a 2 s Gauss pulse train with the irradiation frequency at −0.7 ppm or −50 ppm alternatively. The total acquisition time was 15 minutes with 256 scans.

Screen Analysis: One of the first things they noticed was that there were difference between the reference spectra (plain water) and the screening sample (protein buffer).  They decided they could not automate the entire process and instead just scripted the data processing and display.  Then they confirmed each putative active as a singleton. 

What they are putting together is a "One Size Fits All" process.  I give them credit for doing this, but I think that you cannot find a single NMR-based process for all targets.  In particular, I think they could have used more typical conditions for the reference spectra.  The paper then goes on and discusses their application to targets of interest.  For me, that is irrelevant.  This paper is an excellent companion to the Current Protocol paper, and due to open access, most likely to get far more citations.

Off-rate screening (ORS)

Molecules that dissociate slowly from their target proteins are potentially useful because they can have a long-lasting effect even if they are rapidly cleared from circulation. However, it is next to impossible to predict whether a molecule will dissociate slowly or not. Moreover, the correlation with binding affinity is poor: weak binders generally don’t stay bound to their target for long, but even tight binders often rapidly dissociate. In the early stages of lead discovery most folk are focused on affinity, and it is usually only much later that kinetics enters in. In a new paper in J. Med. Chem., James Murray, Paul Brough, and colleagues at Vernalis introduce a technique that moves kinetics to the front of the line.

The technique, off-rate screening (ORS), relies on surface plasmon resonance (SPR), which is already commonly used to study binding kinetics. The trick here is using SPR to screen products in unpurified reaction mixtures. An initial fragment with known affinity is modified, and products screened for slower dissociation. Of course, the concentration of desired compound is likely to vary from mixture to mixture, but the great thing about looking at compound dissociation is that it is a zero order reaction: it does not depend on concentration. The researchers use mathematical simulations to show that even if the yield is only 5%, a product with a 10-fold slower dissociation rate constant could still be detected. Since off-rates can vary by orders of magnitude, this is not such a high bar.

Of course, simulations are one thing, but how does the technique actually work in practice? The researchers show examples on two targets, one using some of the early compounds for their HSP90 program, the other some of their PIN1 inhibitors. For PIN1, the researchers resynthesized some of the molecules in plastic tubes, which caused leaching of plastic into the reaction mixtures. Nonetheless, for both proteins the dissociation rate constants measured for unpurified reactions were very close to purified molecules, generally differing by less than 30%.

The researchers also tried subjecting compounds to eleven reaction conditions typically used in medicinal chemistry, evaporating the solvent, and testing the products; the idea was to see if the reagents or other components in the reaction mixture would interfere with the assay. Happily in all cases the dissociation rate constants differed by less than 20%, again pointing to the robustness of ORS.

Of course, as with any technique, there are limitations. Since the screening compounds are not purified from their starting materials, the desired products must dissociate sufficiently slowly from the protein to be distinguishable from other components in the reaction mixture; dissociation rate constants greater than about 1.2 s^-1 appear to be challenging. Also, if the starting material itself has a slow dissociation rate from the protein, it may be difficult to differentiate this from a low yield of slowly dissociating product. The researchers note that both cases could be addressed by changing the temperature, either lowering it to slow the dissociation rate constant or raising it to increase it.

All in all this is a nice approach, and it will be interesting to see how widely it catches on.

Flexible fragment linking vs the transcription factor EthR

Transcription factors have a well-earned reputation for being extremely difficult targets. Although the literature is littered with inhibitors of various transcription factors, most of these turn out to be of questionable validity, to put it politely. A recent paper in Biochem. J. by Sachin Surade, Tom Blundell, and collaborators at the University of Cambridge and the Ecole Polytechnique Fédérale de Lausanne-EPFL reports what looks to be the real deal.

The researchers were interested in a protein called EthR, a transcription factor from Mycobacterium tubercuolosis involved in antibiotic resistance. Unlike many other transcription factors, this one contains an allosteric binding pocket known to bind lipophilic molecules. Armed with this knowledge, the researchers performed a thermal shift assay using a library of 1250 fragments at 10 mM each, which resulted in 86 hits that stabilized the protein by at least 1 °C. These were then tested for their ability to disrupt the interaction between EthR and DNA using surface plasmon resonance (SPR), and 45 of them showed greater than 10% inhibition at 0.5 mM. Reassuringly, only 1 of 45 fragments that had shown no stabilization in the thermal shift assay showed more than 10% inhibition here, suggesting that the thermal shift assay had a low false negative rate.

Confirmed hits were characterized by full dose-response curves and soaked into crystals of EthR, resulting in several co-crystal structures. Compound 1 was particularly interesting because two copies of it bound to the central hydrophobic channel, which was only possible due to conformational changes in the protein. Also, although the likely natural ligand of EthR appears to make only hydrophobic contacts to the protein, the carbonyl of compound 1 makes hydrogen bonds. In one of the two bound molecules, the interaction is with an asparagine residue of EthR; in the other, it is with a water molecule.

Swapping the cyclopentyl ring to a phenyl to yield compound 5 gave a slight loss in potency but simplifies subsequent modifications, and crystallography revealed that it binds in the same manner as compound 1. More significantly, linking two molecules of compound 5 via a disulfide bond (compound 9) improved the affinity by more than two orders of magnitude.

Of course, disulfides can react with cysteine residues in a protein – a fact that can be rather useful for finding inhibitors. Thus, it was essential to demonstrate that compound 9 was really binding non-covalently to the protein rather than acting through an unrelated mechanism. Happily, the researchers were able to determine the co-crystal structure of compound 9 bound to EthR, confirming that it binds in the same manner as the two molecules of compound 5, including the two hydrogen bonds. (Unfortunately though, none of the crystal structures appear to be deposited in the protein data bank.)

Compounds 1 and 9 were both tested for their activity to enhance the effect of the antibiotic ethionamide in Mycobacterium tubercuolosis cultures, and both were active, though with similar potencies despite their very different affinities to the isolated protein; it seems likely that the disulfide bond would be reduced in the bacterium. It will be interesting to replace this with a more stable linkage (amides were also tried but did not improve affinity).

One interesting conclusion is that “flexible fragments in the library can lead to a more efficient exploration of chemical space.” This is exemplified by the fact that floppy fragment 1 binds in two somewhat different conformations to the two sites on the protein. Having some flexibility in the early stage of a project can be useful, and another reason not to be too rigid in assembling a fragment library.