27 April 2015

Tenth Annual Fragment-based Drug Discovery Meeting

Last week marked the tenth anniversary of CHI’s three-day Drug Discovery Chemistry conference in San Diego. The conference consists of six tracks, with three happening simultaneously. The FBDD track is the only one which dates all the way back to the beginning in 2006. In fact, this is the oldest recurring fragment conference, predating both the Royal Society Fragments meetings as well as the independent FBLD meetings.

It’s worth reflecting on how far fragments have come since 2006. Back then, as Rod Hubbard (Vernalis and University of York) noted, most of the talks were prospective and methodological. Even as late as 2010 there were talks describing how dedicated fragment groups needed to be shielded from the larger organization. Now fragments are mainstream: a large fraction of the talks in the protein-protein interaction track involved fragments, as did both plenary keynote addresses to the entire conference.

Harren Jhoti’s keynote focused on lessons learned at Astex over the past 15 years. There has been some debate in the literature over ligand efficiency (LE), and one slide that struck me was a summary of 782 dissociation constants (measured by ITC) against 20 projects. The vast majority of these compounds had LE > 0.3 kcal/mol/atom. Given that Astex has put multiple fragment-derived drugs into the clinic and was acquired by Otsuka in one of the largest M&A events of 2013, the metric appears to have some utility.

Still, it’s important not to be dogmatic, particularly for difficult targets. Harren described a program for XIAP/cIAP where they started with an extremely weak fragment with LE < 0.2, but its binding mode was sufficiently interesting that they were willing to work on it. This program also revealed the importance of biophysical measurements, as functional activity was uninterpretable and even misleading until higher affinity compounds were discovered.

One theme throughout the conference was the observation that fragments bind at multiple sites on proteins. Harren noted that Astex researchers have found fragments bound (crystallographically) to 54 sites on 25 targets – an average of 2.2 sites per target. Some targets are even more site-rich: Joe Patel (AstraZeneca) performed a crystallographic screen on a complex of Ras and SOS and found four binding sites, including one previously discussed here. In this effort, 1200 fragments were screened in pools of 4, and in one case two fragments from the same pool each bound only when they were both present at the same time – each fragment alone showed no binding by NMR or crystallography.

Troy Messick (Wistar) described his work against the EBNA1 protein from Epstein-Barr virus. An HTS screen of 600,000 compounds came up with at best marginal hits, but soaking 100 different Maybridge fragments into protein crystals led to 20 structures, with fragments bound to four different sites. Some of these fragments were then merged to give cell-active compounds with good oral bioavailability.

Rather than exploring different ligands binding at different sites, Ravi Kurumbail (Pfizer) described an interesting case of the same ligand binding at different sites. A screen against the kinase ITK identified a (large) fragment that could bind both in the adenine binding pocket as well as a nearby pocket, as determined crystallographically. Determining the affinities of the same fragment for the two sites necessitated some clever SPR and enzymology, but did lead to a highly selective series.

In terms of targets, BCL-family proteins were certainly well-represented, featuring heavily in talks by Chudi Ndubaku (Genentech, selective Bcl-xL inhibitors), Mike Serrano-Wu (Broad Institute, MCL-1 inhibitors), Zaneta Nikolovska-Coleska (University of Michigan, MCL-1), Roman Manetsch (Northeastern, Bcl-xL and MCL-1), and Andrew Petros (AbbVie, BCL-2 and MCL-1). Of course, it was AbbVie (neé Abbott) that pioneered BCL inhibitors as well as FBLD in general, and I was happy to hear that there is a renaissance occurring there, with fragment approaches being applied to all targets, even those undergoing HTS.

Finally, there were some interesting practical lessons on library design. Peter Kutchukian described how the Merck fragment library was rebuilt to incorporate more attractive molecules that chemists would be excited to pursue. There is an ongoing debate as to whether a fragment library should be maximally diverse or contain related compounds to provide some SAR directly out of the screen, and in the case of the Merck library the decision was to target roughly five analogs in the primary library, with a secondary set of available fragments for follow-up studies.

The utility of having related fragments in a library was illustrated in a talk by Mark Hixon (Takeda) about their COMT program. A HTS screen had failed, and even a screen of 11,000 fragments came up with only 3 hits (with an additional close analog found by catalog screening). Remarkably, all of these are extremely closely related, but other analogs in the library didn’t show up; had they not had multiple representatives of this chemotype in their library they would have come up empty-handed.

In the interest of space I’ll close here. Teddy will post his thoughts later this week, and please share your own. CHI has announced that next year’s meeting will be held in San Diego the week of April 19. And there are still several great events on the calendar for this year!

20 April 2015

Tethering versus RNA

Last week we highlighted one of the less common fragment-finding methods, and today we turn to another. Tethering uses reversible disulfide exchange chemistry to trap thiol-containing fragments near binding sites. Back when we developed the technology at Sunesis we used cysteine residues in proteins. We occasionally discussed applying it to nucleic acids, but at the time it was hard to make a good business case. Now that microRNAs (miRNAs) have become hot, there is more interest in going after nucleic acid targets, and in a recent paper in Molecules Kiet Tran and Peter Beal (UC Davis) and Michelle Arkin (UC San Francisco) have done just that.

The researchers were interested in an RNA sequence that is cleaved in cells to generate miR-21, a potential cancer target. The idea is to find small molecules that bind to pre-miR-21 and prevent its processing to the mature miRNA. To perform Tethering, the researchers first introduced a thiol group into adenosine and incorporated this into RNA. They made two separate versions of pre-miR-21, with the modified adenosine at a different site in each, and also made a control RNA with a completely different sequence.

Next, they incubated the modified RNAs with 30 different disulfide-containing small molecules under partially reducing conditions and used mass spectrometry to identify those that covalently bound. As expected most showed minimal binding, but there were a couple hits. One of these, a 2-phenylquinoline, bound to both modified versions of the pre-mR-21 as well the control RNA, suggesting non-specific binding. In fact, 2-phenylquinoline is a known intercalator, so while its identification is not surprising, it does validate the ability of Tethering to identify binders. The other hit, however, appeared to be specific for one of the two pre-mR-21 sequences.

Of course, there is still a long way to go; it is unclear how much affinity the hit has for the RNA, or how specific it would prove if tested against a large panel of decoy RNAs. A key challenge for Tethering – as with many fragment-finding methods – is figuring out what to do with a hit. This is all the more true with RNA, about which we’ve written several times over the years. Still, one nice feature of Tethering is that it allows one to target a specific site of interest. Also, the covalent (disulfide) bond helps with both crystallography and modeling. It will be fun to watch this story develop.

13 April 2015

Substrate activity screening for irreversible PAD3 inhibitors

Of all the ways to find fragments, one of the more unusual is substrate activity screening, or SAS, which we first discussed here. The idea is to make and screen libraries of potential enzyme substrates and transform the best ones into inhibitors. In a new paper in J. Am. Chem. Soc., Jon Ellman and coworkers at Yale University describe how they used SAS to discover irreversible inhibitors of protein arginine deiminase 3 (PAD3), a potential target for neurodegenerative diseases.

The four human PADs (conveniently named PAD1-4) transform arginine residues in proteins to citrulline residues, with subtypes distributed differently across different tissues. The researchers started by making a library of more than 200 fragment-sized guanidines (the unique side-chain moiety in arginine) as potential substrates. These were then screened in a colorimetric assay. Several compounds were found to be processed by the enzyme, though all were very weak substrates (Km > 10 mM).

Next, the best substrates from three different chemical series were optimized for activity. For example, substrate 4a was grown to substrate 15a.


Finally, the substrates were converted to irreversible inhibitors by replacing the guanidine with a known chloroacetamide warhead. This coopts the natural mechanism of the enzyme, which relies on covalent bond formation between an active-site cysteine residue and the substrate. Within a given series, the better the substrate, the better the resulting inhibitor (for example, inhibitor 15b is more potent than inhibitor 4b). However, these correlations did not hold across series.

The best compounds were also tested for selectivity, and some of them were at least 10-fold selective for PAD3 over the other three PADs.

Last year we highlighted a paper that described several difficulties encountered (and overcome) using SAS to find inhibitors of the protease urokinase. (The comments to that post are well worth reading as they include contributions from the corresponding author of the paper as well as a former Ellman postdoc who is using SAS.) However, according to the current paper, SAS was relatively straightforward for PAD3 – another confirmation that different targets require different approaches.

08 April 2015

Fragment-based methods in drug discovery

FBLD generates a plethora of reviews, as evidenced by Practical Fragments’ annual round-ups (see for example 2014, 2013, and 2012). However, for the past three years there have been no new books. The drought has now ended, starting with the publication of Methods in Molecular Biology Volume 1289, edited by Anthony E. Klon of Pennsylvania Drug Discovery Institute. Computational chemistry is probably one of the most rapidly changing disciplines within FBLD, and thus it is appropriate that this is the primary focus.

The book is part of the Springer Protocols series, which offers highly specific step-by-step instructions. Many of the chapters have a common organization: Introduction, Materials, Methods, and Notes. While this can work well for established molecular biology techniques such as cloning, it can be trickier to apply to computational approaches. Some of the chapters are quite brief and assume extensive specialized knowledge, while others are extremely detailed. Of course, it is impossible to satisfy everyone; hopefully the following summary will help you find what is most useful for you.

Part I (Preparation) consists of five short chapters. The first is by Rachelle Bienstock, editor of the most recent (and also computationally intensive) book. As we’ve noted, water plays a pivotal role in protein-ligand interactions, and Rachelle concisely but thoroughly summarizes available computational methods. Chapter 2, by Yu Zhou and Niu Huang at the National Institute of Biological Sciences in Beijing, outlines how to use DOCK to assess binding site druggability. In chapter 3, Raed Khashan (King Faisal University, Saudi Arabia) describes a free software tool called FragVLib for generating virtual fragment libraries to compare different ligand binding pockets. Chapter 4, by Jennifer Ludington (formerly of Locus Pharmaceuticals), discusses practical issues in preparing a virtual fragment library, such as conformer and partial charge assignment. Finally, in chapter 5 Peter Kutchukian discusses how he and his Merck colleagues enlisted medicinal chemists to help fill the gaps in their fragment collection.

The second section is titled Simulation. In chapter 6, Kevin Teuscher and Haitao Ji (University of Utah) summarize “fragment hopping,” including an extensive table of available software tools. Chapter 7, by Olgun Guvench (University of New England), Alexander MacKerrel (University of Maryland), and coworkers describes SILCS: site identification by ligand competitive saturation. This program, developed by SilcsBio LLC, soaks proteins in virtual solutions containing very tiny fragments (think propane and methanol) to look for binding sites. Molecular dynamics simulations include methods to prevent aggregation of the ligands or denaturation of the protein.

Chapter 8, by Álvaro Cortés-Cabrera, Federico Gago (Universidad de Alcalá, Madrid) and Antonio Morreale (Repsol Technology Center, Madrid), describes how ligand efficiency indices can be used to guide fragment growing. Of course, metric skeptics will still ask, “sure it works in practice, but does it work in theory?” And in chapter 9, Jui-Chih Wang and Jung-Hsin Lin (Academia Sinica, Taipei) introduce a new scoring function for fragment-docking, including several pages of detailed instructions for implementing it in AutoDock. As we’ve noted, calculating binding affinities for fragments can be difficult, and the new function seems to be accurate to about ±2.1 kcal/mol for a series of compounds tested

Part III, Design, begins with another chapter by Rachelle Bienstock in which she outlines the process of fragment-based ligand design, highlighting various software tools available at each stage. This includes library design, growing, linking, and downstream considerations such as ADME. Chapter 11, by Zenon Konteatis of Agios, is a brief primer of the process, including an example for the kinase TGF-beta. The last chapter in this section, by Jennifer Ludington, focuses on binding site analysis to assess whether a protein site is druggable (or at least ligandable). She focuses on the procedure used at Locus Pharmaceuticals, which involved soaking a virtual protein in a solution containing fragments and then lowering the chemical potential of the system until only the tightest fragments remain bound. Clusters of probe fragments indicate possible hot spots.

Finally, Part IV consists of Case Studies, starting with a chapter on kinase inhibitors by Jon Erickson (Lilly). More than a third of FBLD-derived clinical candidates target kinases, so it is always good to have an updated overview, though there is at least one structural error.

The last two chapters are both by Frank Guarnieri, founder of Locus Pharmaceuticals and currently at Virginia Commonwealth University School of Medicine. These are highly opinionated (with lots of first-person singular pronouns) and fun to read. They both describe the simulated annealing of chemical potential (SACP) approach that formed the basis of Locus (and is also discussed by Jennifer Ludington above). Chapter 14 describes a small molecule erythropoietin (EPO) mimetic program. The protein EPO binds to and activates a dimeric receptor, and a small molecule functional mimetic would indeed be an exciting breakthrough. Unfortunately, the primary data presented are not compelling, and I remain unpersuaded, though perhaps readers are aware of more convincing evidence.

Chapter 15 describes the Locus program to develop a highly selective orally available p38 inhibitor. The discussion offers a rare window into life at a small biotech, including disagreements over strategies and interpretation of data. It now appears that p38 is probably not a good target for inflammation, which had unfortunate repercussions:

The business decision at Locus to put so many resources into this program along with other questionable business decisions resulted in the company going bankrupt after about 10 years in existence.

Some of the most important lessons are negative, and it’s nice to see these appear in print. Success stories are inspirational, but this chapter is a healthy reminder of the very many things that must succeed for fragment-based approaches to yield new drugs.

06 April 2015

When a Lead is a Lead

As we keep on saying, epigenetics is big.  So, today we present another paper on an old friend, BRD4.  This paper is a follow up from previous work where they used docking and X-ray to find the thiazolidinone fragment hit that was elaborated as shown below (Figure 1), but with potency in the single digit micromolar in vitro and double digit in cellulo
Figure 1.  Previous work from these authors.
In this work, they continue developing this scaffold investigating the reversed sulfonamide(Figure 2)
Figure 2.  Reversed Sulfonamide
which had significantly improved activity.  Cyclo-aliphatic rings showed increases in potency, but with larger rings also decreasing ligand efficiency.  Aromatic rings decreased potency and larger groups (rings with linkers) were not tolerated at all.  

The crystal structure of the cyclopentyl derivative was solved and was seen to have a different binding mode from the original fragment.  In this case, the WPF shelf is NOT the major binding site for the compound.  In the end, they ended up with
Figure 3.  End Result of this study.  
This is compound is potent (albeit not super potent), ligand efficient, with cell-based activity,  selectivity, and good PK properties.   What I really like is that final sentence of the conclusion:
a promising BRD4inhibitor and a useful lead for further anticancer drug development.

01 April 2015

Shapely fragments

Tired of all those planar aromatics in your compound collection? Three-dimensional fragments are all the rage these days, and chemical suppliers are happy to oblige. After the stunning success of their FUNK library, SerpentesOleum has come out with a new offering, Tesseract Products (TP). All of the TP fragments are guaranteed to be nice, plump, and squeezably soft. For example:


Even more exciting, the company has hired a crack team of physicists to produce a line of 4-dimensional fragments with principal moments of inertia greater than 1. Don't delay, order your TP today, and wipe away the 2-D blues!

30 March 2015

Politburo Approved

Viral, tropical diseases are really cool because they have great names. e.g. Dengue or Breakbone Fever or Chikungunya ("that which breaks up").  The great thing about many viral diseases is that they are dependent upon proteases for many things. (And yes, I know how that sounds.)  Proteases have nice, well defined active sites that you can fill quite well and shut them down. In this paper, the authors use fragment-peptide merging to inhibit Dengue protease.  

This is really an extension of previous work.  The original work used capped peptides with a warhead with very good potency (down to 43 nM).  They then investigated retro, retro-inverse, semiretro-inverse, and nonretro di- and tri-peptides.  This lead them to use a tri-peptide (Arg-Lys-Nle) in two generations: first an arylcyanoacrylamide and then to N-substituted 5-arylidenethiazolidinone (thiazolidinediones and rhodanines).  These second generation hybrids had increased membrane permeability, in vitro binding, in cellulo antiviral activity.  Based on docking, they decided to investigate Nle sitting in P1', in contrast to previous site preferences and then merge it with fragments from an optimized capping moiety. 
1.  Starting Point Hybrid Peptide
The investigation of Nle replacements led to the phenylglycine molecule, with 4x greater affinity:
9.  Phenyl-glycine hybrid
They, then chose three hybrids (including 9) and put two different caps on them:
Rhodanine Cap
Acrylamide cap

Compared to the benzoyl cap, the acrylamide was 2x better while the rhodanine was 5x better.  But, wait, doesn't the Politburo condemn all uses of rhodanines?  Of course not.  In this case, the rhodanine was selected through rigorous analysis: and they have selectivity (this assay is fluorogenic).  They are perfectly aware of the general distaste people have for rhodanines and address the concerns. All of this together, leads to the final compound (below).
This is a really nice piece of starting with a tool (covalent peptides) and working to generate drug like molecules with favorable properties. 

25 March 2015

We read these papers so you don't have to

Glycogen Phosphorylase is one of those systems that you hear about all the time; it was the first allosteric enzyme discovered.  It's been discussed here and here previously on this blog.  It is one of those ubiquitous enzymes and has been the subjet of a lot of research looking for allosteric modulators.  The majority of allosteric inhibitors are heterocyclic compounds with a well known history.  This paper wants to add to that history. 
The authors start with what appears to be a dreadful understanding of what fragment-based hit generation is.
"Lead-like discovery refers to the screening of low molecular weight libraries with detection of weak affinities in the high micromolar to millimolar range".
Maybe its just me, but we've been over this before.  Lead-like molecules, as Kubinyi showed, are large and decorated; fragments are not.  So, they got the low molecular weight thing right, but the name of the method wrong. Maybe an error in the proofing...
Starting on previous work, the chose a 21 member heterocycle library (Figure 1.) to investigate a morpholine-based peptide mimetic.
Figure 1.  Fragment Library
Activity was determined by an enzymatic assay with a maximal compound concentration of 222mM.  They also used 22mM, 56 mM, 111mM leading to Table 1 and some crazy SAR (N-Boc-ing 8 yielded 9 with >200x potency).  
Table 1. 
The key compound is 7, with 25 microM IC50;  while 6 (minus the methyl ester) is 1000x less potent.  Strange things are afoot at the Circle K.  They then docked 7 (and a few other "second tier" compounds).  They see "moderate" binding for all compounds; yet, one of these compounds is more than 50x more potent than the others.  We've been down this road before...

23 March 2015

Rad fragments revisited

Two years ago we highlighted a paper in which Cambridge University researchers identified fragments that bind to the protein RAD51, which in turn binds to the protein BRCA2 to protect tumor cells from radiation and chemotherapeutics. In a new paper in ChemMedChem, Marko Hyvönen and colleagues describe how they have grown these fragments into low micromolar binders.

One of the best fragments identified in the previous work was L-tryptophan methyl ester (compound 1), so the researchers naturally tried substituting the methyl group. A phenethyl ester (compound 5c) gave a satisfying 10-fold boost in potency, but this turned out to be the best they could get: shorter or longer linkers were both less active, and modifications around the phenyl ring gave marginal improvements at best. Also, changing the ester to an amide decreased affinity. They were, however, able to improve potency another order of magnitude by acylating the nitrogen (compound 6a).


At the same time, the researchers made a more radical change to the initial fragment by keeping the indole and replacing the rest with a sulfonamide (compound 7a). This also boosted potency. Further optimization of the sulfonamide substituent improved the affinity to low micromolar (compound 7m) and increased ligand efficiency as well.

The original fragments had been characterized crystallographically bound to the protein, but the researchers were unable to obtain structures of the more potent molecules, though they did sometimes see tantalizing hints of electron density. Competition studies with known peptide inhibitors also suggested that the molecules do bind in the same site as the initial fragments.

The thermodynamics of binding were characterized using isothermal titration calorimetry (ITC). Although the initial fragments owed their affinity largely to enthalpic interactions, the more potent molecules were more entropically driven. This, the researchers suggest, could partially account for the failure of crystallography despite extensive efforts: the lipophlic molecules can bind in a variety of conformations.

Some have argued that enthalpic binders should be prioritized, but this study illustrates one of several problems: even if you start with an enthalpic binder, there’s no guarantee it will stay that way during optimization.

This is a nice paper, but I do wonder how much affinity there is to be had at this site on RAD51. Given the micromolar affinity of the natural peptides, nanomolar small-molecule inhibitors may not be possible. Then again, like other difficult PPIs such as MCL-1, perhaps the right molecule just hasn’t been made. How long – and how hard – should you try?

18 March 2015

Mass Spec Screening in Solution

Mass spectrometry is a technique that most people are familiar with, as a QC tool.  It also has been demonstrated as a screening/validation tool.  Native mass spectrometry (nMS) has been discussed here, Weak Affinity Chromatography (WAC) here, and Hydrogen-deuterium exchange (HDX) here.  All of these methods have advantages and disadvantages.  A "new" method is the ligand-observed MS screening (LO-MS).  [I put new in quotes because I know of at least one company that has been using this method for screening for years via a CRO.]

The concept of LO-MS is straight forward (Figure 1) and very similar to WAC.  A mixture of fragments, in this case 384, are mixed with target (NS5B), incubated, and the ultrafiltrated (50kDa cutoff).  This step eliminates the need for the immobilization step in WAC, ensuring the native conformation.  The fragments were at 25 uM, while the target was at 50 uM. 
Figure 1.  Fragments MW 165 and 130 are binders.  MW162 and 150 are not. 
Retained fragments are then dissociated with 90% methanol and those showing intensity higher than the protein-minus control are considered binders (S/N  greater than 10).  In their library, 5% of the compounds were not amenable to mass spec detection, but they included them to increase the complexity of the mixture.  In the end, they ended up with 20 binders in 20 minutes!  They repeated the screen with smaller mixtures (50 and 84 fragments) where they found 12 binders (a subset of the original 20).  As a follow up, they ran the binders by SPR, validating 10 of the binders (50%).  5 out of these 10 gave useable crystals (observable electron density for the fragment) (50%).  They also show how the data can be used to generate Kds (like WAC).

This method raises some issues with me, but first let me say, it sure seems to work, and fast to boot.  From people I know who have used this to screen, they have been very happy.  Here is what bothers me: self-competition in the tube a discussed here and here, this is a non-equilibrium method (variable protein concentration during the ultrafiltration), and it is an indirect method.  For me, I prefer methods that directly detect ligand-target interactions, like NMR, SPR, and nMS.

16 March 2015

Fragments vs p97

The protein p97 is important in regulating protein homeostasis, and thus a potential anti-cancer target. But this is no low-hanging fruit: the protein has three domains and assembles into a hexamer. Two domains, D1 and D2, are ATPases. The third (N) domain binds to other proteins in the cell. All the domains are dynamic and interdependent. Oh, and crystallography is tough. Previous efforts have identified inhibitors of the D2 domain, but not the others. Not to be put off by difficult challenges, a group of researchers at the University of California San Francisco (UCSF) led by Michelle Arkin and Mark Kelly have performed fragment screening against the D1 and N domains, and report their adventures in J. Biomol. Screen.

Within UCSF, the Small Molecule Discovery Center (SMDC) has assembled a fragment library of 2485 commercial compounds from Life, Maybridge, and Asinex. These have an average molecular weight of 207 Da and 15 heavy atoms, with ClogP ~1.5. The researchers used both biophysical and virtual screening.

For the physical screening, the researchers started with surface plasmon resonance (SPR), with each fragment at 0.25 mM. This resulted in 228 primary hits – a fairly high hit rate. Full dose response studies revealed that 160 of theses fragments showed pathological behavior such as concentration-dependent aggregation or superstoichiometric binding. A further 30 showed weak or no binding, 13 were irreversible, and 5 bound nonspecifically to the reference surface, leaving only 20 validated hits which were then repurchased.

The 228 primary hits were also assessed by STD NMR, each at 0.5 mM when possible (some fragments were not sufficiently soluble). Of these, 84 gave a strong STD signal, and 14 of these were also among the 20 SPR-validated hits.

The 20 repurchased fragments were further tested by both SPR and STD NMR, and 13 of them reconfirmed by both methods. The paper includes a table listing all 20 compounds, and one observation that struck me was the fact that all but one of the hits – which had dissociation constants ranging from 0.14 to 1.7 mM – are larger than the library average. Such results could argue for including larger fragments in libraries, though this goes against both molecular complexity theory as well as extensive experience at groups such as Astex.

Next, the researchers sought to discover information on the binding sites. Three fragments could be competed by ADP, suggesting that they bind in the nucleotide-binding site of D1. To narrow things down further, the researchers turned to 13C-1H-methyl-TROSY NMR, in which specific side chain methyl groups of Ile, Leu, Met, Val, and Ala were labeled, and chemical shifts were examined in the presence and absence of fragments. Two of the proposed nucleotide-binding site fragments showed similar shifts as AMP or ADP, further supporting a common binding mode (the third was too weak to test). This was not an easy experiment: the hexamer has a mass of 324 kDa, well above where most people do protein-detected NMR.

Independent of all the biophysical screens, virtual screens were conducted using Glide XP, which suggested that the nucleotide binding site would be the hottest hot spot. Happily, all three fragments that appear to bind to this site scored highly in the in silico work, with two of these within the top 100 fragments. However, the binding sites for the other ten confirmed fragments remain obscure.

This paper serves as a useful guide for how fragment screening is performed on a tough target in a top-tier research group. Although difficult, it is not impossible to advance fragments in the absence of structure. While it remains to be seen whether that will be the case for any of these fragments, the researchers have provided a wealth of data for those who wish to try.

11 March 2015

The Sequel is Never as Good as the Original

We are living in a target-driven environment in Pharma, for both good and bad.  The low-hanging fruit have been plucked and the high-hangers are tough.  But, fragments have proven to be highly utile in liganding these targets.  One drawback with target-based screening is the problem with cellular activity, while it may be easy to generate good activity against the isolated target, in the end you need activity in the cell/animal.  Back in the good ole days, people just skipped the target and went straight into cells: compounds are put on bacterial plates and the microbes die if the compound is anti-microbial.  This is the simplest example of phenotypic screening, the phenotype here being "dead cells". [For a discussion of the history of phenotypic screening, go here.]  Fragments could be the worst case scenario for phenotypic screening as fragment-target interactions are very weak, and very commonly do not exert a biological effect. 

In this paper from Rob Leurs and colleagues, including Iota, the describe a fragment-based phenotypic screen process.  This work is a follow on to previous work from this group discussed here, which I quite liked  So, they have a target (PDEB1) but immediately follow their screening with the phenotypic part.  For the phenotypic screen, they used several different parasitic PDE and MRC5 cell-line as a counter-screen. I won't bore you with any of the experimental details. The compounds are recapitulating known molecules, like benadryl.  Now, I really wanted to like this paper, at least from a process approach.  It appears to my eyes, that all the compounds are pretty much equipotent and cytotoxic.  This is a really disappointing paper in that it doesn't really do anything.  They had shown previously that you could get non-cytotoxic compounds with good inhibition of PDEB1.  They didn't repeat that here.  There is no X-ray, they did before.  The compounds are wholly uninteresting and stretch the imagination to be seen as compounds "with a lot of potential to grow into antiparasitic compounds".

09 March 2015

Are PrATs privileged or pathological?

Pan assay interference compounds – PAINS – have received quite a bit of attention at Practical Fragments. In addition to being a fun topic, the hope is that publicizing them will allow researchers to recognize them before wasting precious resources.

But not all PAINS are created equal. Some, like toxoflavin, simply do not belong in screening libraries due to their tendency to generate reactive oxygen species. I would put alkylidene rhodanines in the same category due to their ability to act as Michael acceptors, their tendency to undergo photochemistry, and their hydrolytic instability. The nice thing about these sorts of molecules is that their clear mechanistic liabilities justify excluding them.

But things are not always so simple, and in a recent paper in J. Med. Chem. Martin Scanlon and co-workers at Monash University, along with J. Willem Nissink at AstraZeneca, describe their experiences with a more ambiguous member of the PAINS tribe: 2-aminothiazoles. (See here for In the Pipeline’s discussion of this paper.)

That 2-aminothiazoles (2-ATs) should be PAINS is not obvious: at least 18 approved drugs contain the substructure. Thus, it was not unreasonable to include 2-ATs in the 1137-fragment library assembled at Monash. But after screening 14 targets by STD-NMR and finding a 2-AT hit in every campaign, the researchers started to become suspicious. They gathered a set of 28 different 2-ATs and screened these against six structurally diverse proteins using surface plasmon resonance (SPR). Many of the 2-ATs bound to 5 of the proteins, and a couple bound to all six. The researchers used 2D-NMR (HSQC-NMR) to further characterize binding and found that the 2-ATs bind to multiple sites on the proteins rather than the desired one-to-one binding mode.

A common source of artifacts is the presence of reactive impurities, so the researchers resynthesized some of the 2-ATs and showed they behave the same, ruling out this mechanism. Solubility was also not a problem. Finally, the ligand-based NMR experiments revealed that the 2-ATs really did appear to be binding to the proteins, ruling out interference from unreacted starting materials or decomposition products.

One structure-activity relationship did emerge: acylation of the amino group dramatically reduced promiscuity of the 2-ATs. However, in the case of 2-ATs with a free amino group, there was little meaningful SAR. Thus, the researchers propose calling these molecules PrATs, or promiscuous 2-aminothiaozles.

Further analysis of high-throughput screening data from the Walter and Eliza Hall Institute and AstraZeneca revealed that 2-ATs were also over-represented among hits. What’s spooky about this result is that most of the screens were done at 10 micromolar – far lower than typical fragment screens.

The researchers freely admit that they have no mechanism for why PrATs bind to so many proteins. I suspect there is something fundamental to be learned about intermolecular interactions here, though how to extract these lessons is beyond me. One gets the impression that the authors themselves have been burned by pursuing PrATs, as they conclude:
On the basis of our findings reported here and our unsuccessful attempts to optimize these fragments against different targets, we have removed 2-ATs from the fragment library.
This paper serves as a thorough, cautionary analysis. As evidenced by multiple approved drugs, PrATs can be advanceable, and we certainly won’t be PAINS-shaming papers that report them as screening hits. If you can advance one to a potent lead, then bless your heart. But be warned that this is likely to be even more difficult than normal.

04 March 2015

Way Down in the X-Ray Weeds

So, what I know about the details of crystallography can fit on the head of a pin...a small pin.  You put pure protein in multiwell plates and then do a huge matrix of crystallization conditions until tiny little crystals form.  Big crystals are best, but you can use tiny crystals or seeds, or with recent advances in technology, to actually collect data.  Then, through some wizardry (some sort of inverse transform) you make spots go to electron density, then with will power and what used to be SGI machines, you thread your protein sequence in, et voila a model of the structure.  I typically don't go for methods papers in fields I have almost no clue in, but this one intrigued me.

This paper aims to increase the efficiency of soaking fragments into crystals to take advantage of 3rd generation synchrotrons. These machines/labs/setups/doohickeys use acoustic droplet injection (ADE), which many people may already be aware of.  In this approach, each fragment soaks into a protein crystal either directly on data collection media or on a moving conveyor belt which then delivers the crystals to the X-ray beam.  The source of inefficiency comes from the time required to soak the fragment in to the crystals (for those where the apparatus is inside the X-ray station. I have no idea what that means, but here is a google image search that might give you an idea.)  A second source is the limit of evaporative dehydration during the fragment soak.  

Using the model system lysozyme and thermolysin the identified factors which can increase efficiency: namely smaller crystals can be used to decrease the soak time.  By small crystals, they are talking things that are 100 microns or less.  The authors go on to state that:
These techniques efficiently use fragment chemicals (~2.5 nL per screened condition), protein (~25 nL per screened condition), space (1120 screened conditions per standard shipping Dewar; no limits using a conveyor belt), and synchrotron beam time (less than 1 second/screened condition).  Evaporative dehydration of the protein crystal limits these fragment screening applications to systems where the fragment soak time is not prohibitive. Slow-binding compounds can be screened (without time constraint) in trays using ADE, but will consume significantly more resources such as purified protein and chemical compounds (~1 µl per screened condition). Hence, it is desirable to identify promising cases where the cost-efficient on-micromesh or on-conveyor soaking methods are adequate.
So, what did I really get out of this paper.  I am amazed by the miniaturization and automation that exists in synchrotrons.  It is really amazing. Its good to get and read the literature for another field.  It can be enlightening.  If I am looking at this correctly, they can screen a 1000 fragments with ~1mg of protein.  With that said, how many people screen for fragments in this way?  It seems not to be resource intensive, if you are sitting at a synchrotron.  But, how much does it cost to sit at that synchrotron?  Are the problems called out here something people encounter every day, or is this a "First World problem" for those sitting at synchrotrons.

What really got me was that the majority of the authors are high school students and undergraduates. This emphasizes to me the commoditization of X-ray, and really all services.  There is a very high level of training that goes in to solving the structure; I get that.  But it seems that many of the steps are commmodities, if you will.  When I was at Merck, they had a directive called (in some form) I-C-E: Innovation-Commoditization-Experimentation.  The concept was the highly trained (and highly paid) scientists needed to focus on innovation.  Once innovation was achieved it led to experimentation (figuring out how to run it routinely).  After that it was a commodity and should be outsourced to enable those scientists to go back to innovating.  It makes sense from a business standpoint, but scary from the scientists standpoint.  I am all for full employment for scientists in industry (trust me on this), but outsourcing can co-exist in industry. Look at the growth in providers from 2011 to 2014. Not sure where I am going with this, but food for thought.

02 March 2015

Fragments vs Factor XIa

Blood clotting is something we’re all familiar with, but the details are devilishly complex; lots of different proteins play a role. Physiologically this makes sense: the many components make for a finely tuned system, and you want clotting to happen when it needs to and then stop. Start too late and you might bleed to death. Start too early (or don’t stop) and you could develop a fatal clot. Not surprisingly, lots of things can go wrong, and many of the enzymes involved are drug targets. In a paper recently published in PLoS One, Ola Fjellström and colleagues at AstraZeneca describe their efforts on one of these.

Factor XI is involved in the “amplification phase” of coagulation, and the activated form (FXIa) is a potential antithrombotic and profibrinolytic target. A high-throughput screen had failed to find anything useful, so the researchers turned to fragments.

The team started with a computational screen of 65,000 in-house compounds with molecular weights < 250 Da. They used Schrödinger’s Glide software and previously determined crystal structures of the protein. The top 1800 fragments were then tested using ligand-detected NMR in pools of 6, with each fragment present at 0.1 mM. The researchers were trying to avoid strongly basic compounds, and they found 13 hits with calculated pKa< 9. Next, 600 structurally related analogs of the hits were screened, resulting in 50 hits total. These were then triaged using inhibition in solution (an SPR technique described here) and taken into crystallography trials. Two fragments gave high-resolution structures and were prioritized. Satisfyingly, the two fragments bound as had been predicted by the initial virtual screen.

Fragment 5 was particularly interesting because it had never been observed as a hit in the S1 pocket of a serine protease. Many enzymes in the coagulation cascade share a conserved S1 pocket. This has a predilection for highly positively charged species, so the neutrality of this fragment was attractive.
Separately, the team found a Bristol-Myers Squibb patent application describing compound 9, which they made and characterized crystallographically. The structure suggested merging a portion of compound 9 with fragment 5, and the resulting compound 13 turned out to be one of the most potent FXIa inhibitors reported.

To better understand the system, the researchers took a deconstruction approach to compound 9, testing the portion (compound 15) that had been used in the merging. This bit has low affinity by itself. Yet, when linked to fragment 5, the resulting compound 13 binds roughly 200-fold more tightly than simple additivity would predict. Similarly dramatic fragment deconstruction results have been reported previously for the related enzymes factor Xa and thrombin.

Unfortunately compound 13 has fairly low membrane permeability, high efflux, and high clearance in rats, though preliminary SAR suggests this is the fault of the Bristol-Myers Squibb piece rather than the new fragment. At any rate, this is another nice example of using fragment screening to replace one portion of a known molecule with a new fragment.

25 February 2015

New PAINS, and their painful mechanisms

Pan-assay interference compounds – PAINS – are a topic that has come up repeatedly at Practical Fragments (see here, here, and here for starters). Indeed, they form the basis of our occasional (and controversial) “PAINS shaming” series (see here, here, here, and here). In a paper recently published online in J. Med. Chem., HTSPains-master Mike Walters (University of Minnesota) and collaborators at the Mayo Clinic College of Medicine and AstraZeneca characterize several new classes of PAINS (also covered at In the Pipeline). I was honored with the invitation to write a Viewpoint on the topic. Since both papers are open-access I’ll just briefly touch on a few key points here.

First, one of the only critics of the PAINS concept is concerned that PAINS were originally defined on the basis of their over-representation as screening hits in one set of assays. The new paper goes beyond empiricism to characterize mechanisms, which involve non-specific reactions with thiols. (The “non-specific” aspect is key to their undesirability, as covalent drugs can be quite attractive.)

Second, just as “not every clam will hurt you”, not every molecule with a PAINS substructure will show activity in every assay. These “structure-interference relationships” (SIR) can be mistaken for “structure-activity relationships” (SAR), making PAINS all the more insidious. The researchers explore some of the reasons for the observed SIR.

Third, one of the saddest parts of the new paper is a list of dozens of references in the supplemental information in which PAINS were reported as screening hits or probes. It’s a safe bet that most – if not all – of these should be disregarded.

Finally, publishing this list of new PAINS will allow people to steer clear of them. To borrow from Hippocrates: chemical space is big, life is short. Why waste time working with chemotypes known to be pathological? 

23 February 2015

Heptares Get Bought

Yesterday brought news that Heptares was purchased by Sosei.  The deal is for 180$MM upfront and 220$MM in royalties.  Similar to the Astex deal, Sosei is not absorbing Heptares, but is keeping them as a wholly owned subsidiary.  In terms of clinical assets, Heptares didn't show up on the 2015 version of fragments in the clinic, but they do have a p1 asset, a M1 agonist.  As if we needed further validation, this deal shows the power of SBDD/FBDD and gives a good idea off how it is valued.  Congratulations to our friends at Heptares.


18 February 2015

19F Target-based Screening Reduced to Practice

If you read this blog you know I love 19F NMR.  I am a big fan of it for ligand based screening or as a secondary screen in target-based mode.  Well, this paper is the first to use target-based NMR to screen small molecules.  Using 3-fluorotyrosine-labeled protein and using what they call Protein observed fluorine NMR (PrOF-NMR), the interrogate a PPI (CBP/p300-KIX) and determine this interaction's ligandability. 

Starting from the Maybridge Ro3 Library, they found 508 19F containing fragments.  These were put into 85 mixtures (5 or 6 compounds each) and screened at 833uM and 2.5 % DMSO (Figure 1, KG-501 is a control compound).  Each experiment was performed in 5 minutes (with a quick reference spectrum) which is faster than SOFAST HSQC (for 15N labeled proteins).  15 mixtures gave hits which upon deconvolution gave 4 actives (>2 SD in chemical shift), a 0.8% active rate.
Figure 1.  Typical 19F Data
 They then titrated the four confirmed fragments to determine the Kd, which for 3 of them was in the mM range.  They followed this up with Analog by Catalog and developed some SAR.  Lastly, they used H-N HSQC to verify that the compounds do bind and where they data shows.  They do. 

Some thoughts on this.  The use of 3FY is only one amino acid that can be used.  Fluorotryptophan can also be used, so this method can easily be applied to other systems.  Secondly, 19F can accomodate much larger pool sizes (within reason).  And there is no reason why both could not be used.  Of course, one of the things not noted here is that they produce single mutants in order to ID each individual residue.  I think you could live without residue specific assignments and still get tremendous value out of this method.  I would be curious what others think.  We bring up impractical tools all the time, so I really want to applaud this paper.  Here is a very practical new method for screening. 

16 February 2015

MELK part 2: fragment deconstruction

Last week we highlighted a paper from Chris Johnson and collaborators at Astex and Janssen in which they used fragment-growing to develop a selective inhibitor of maternal embryonic leucine zipper kinase (MELK). The paper immediately following in ACS Med. Chem. Lett. also describes the team’s efforts to discover MELK inhibitors, but using a very different approach.

In fact, the second paper doesn’t really start from fragments. The researchers were interested in designing compounds that bind MELK in a particular fashion: type II inhibitors fit into the hinge region but also insert themselves deep into a back pocket which is accessible when the so-called DFG-loop swings open (DFG-out). Since Type II inhibitors make more interactions with the kinase, they have the potential to be more selective.

The problem was that all previous crystal structures of MELK were “DFG-in”, so the researchers couldn’t use crystal soaking. Instead, they turned to structure-based design for molecules that would be able to span the hinge region and the back pocket. Happily, they succeeded with compound 2, and further optimization led to the low nanomolar compound 7. Co-crystallization experiments using a related molecule revealed that the compound binds as expected with MELK in the DFG-out conformation.


Compound 7 was tested in a panel of 243 kinases and inhibited 31 of them >50% at 1 µM; besides MELK, six other kinases were inhibited with IC50 values < 100 nM. This isn’t terrible, but it is far from the selectivity seen with MELK-T1, the Type I inhibitor discussed last week. Thus, one can’t assume that Type II binders will necessarily be more selective than Type I binders.

Fragments enter the picture at the very end of the paper, when the researchers “deconstructed” their molecules. Simply removing the aminomethyl group from compound 2 to give compound 8 reduced affinity by more than ten-fold. This was expected because crystallography had already revealed that this moiety makes electrostatic interactions with aspartate and glutamate residues in the protein.

More surprisingly, removing the phenyl group from compound 8 produced a molecule with greater affinity and ligand efficiency than the initial compound 2! The researchers determined the crystal structure of this (compound 9) bound to MELK and found that, in contrast to the other molecules, it binds in the DFG-in conformation. The isoquinoline hinge binder actually binds in a similar manner as it does for its DFG-out binding cousins, it’s just the back pocket that is cut off. The researchers speculate that the DFG-in conformation of the protein may be lower energy, giving the edge to compounds that bind to this state. Whether or not this is the case, it is certainly another reminder of the remarkable plasticity of proteins.

09 February 2015

Fragments vs MELK part 1: a chemical probe

As we recently noted, there are hundreds of human kinases in the human genome, and figuring out what they do is not easy. Selective small molecule inhibitors can probe an uncharacterized kinase’s function, but these don’t exist for the majority of proteins. Such was the case for maternal embryonic leucine zipper kinase (MELK), a potential anti-cancer target. In a recent paper in ACS Med. Chem. Lett., Chris Johnson and collaborators at Astex and Janssen describe how they created one.

The researchers started with a screen of the ~1500 Astex fragment library using both ligand-observed NMR and protein thermal shift assays. Hits were soaked into crystals of MELK, resulting in “a large number” of structures of fragments bound to the hinge-binding region. It is certainly possible to develop selective inhibitors that bind to the hinge region, but doing so is seldom straightforward, and thus the researchers were particularly interested in unusual fragments. Compound 1 caught their attention because it makes just a single hydrogen bond with the hinge region via the carbonyl oxygen in the fragment; most other hinge binders form two or three hydrogen bonds.


Compound 1 was well-positioned for fragment growing, and the addition of a phenol moiety (compound 2) led to a nice boost in potency and maintenance of ligand efficiency. Crystallography revealed that the phenolic oxygen was both a hydrogen bond donor as well as an acceptor, and replacing this moiety with a pyrazole led to compound 4, with slightly better potency. Pyrazoles themselves are often hinge binders (two of Astex’s clinical compounds, AT9283 and AT7519, contain them), but crystallography revealed that the binding orientation remained the same as in the original hit.

Expanding the aliphatic ring of compound 1 by one methylene gave compound 5, with a higher affinity and a slightly better fit to the protein, and adding bits from compound 4 gave mid-nanomolar compound 7, or MELK-T1. Impressively, the researchers improved the ligand efficiency of their molecules even as they became larger.

Finally, the moment of truth: would optimizing a fragment with an unusual hinge-binder lead to a selective inhibitor? The researchers tested MELK-T1 in a panel of 235 kinases, and happily only 6 were inhibited >50% at 1 µM [compound]. Notably, these did not include AMPK, which has 60% identity to MELK in the kinase domain. MELK-T1 was also cell-permeable.

This is a classic example of FBLD enabled by a robust protein construct suitable for crystal soaking. Getting to MELK-T1 required the synthesis of only ~35 compounds, and should lead to some interesting biology. At the same time, the researchers took a different approach to come up with another series of leads, which will be the subject of my next post.

04 February 2015

Structure based Design on Membrane Proteins

GPCRs are a big target class, which have historically be unamenable to FBDD/SBDD.  However, recent work has changed this thinking.   Membrane proteins are being viewed as increasingly ligandable and amenable to FBDD.  In this paper, Vass and colleagues show their computational approach to indentifying multiple fragment binding sites amenable to linking.  

Recent clinical evidence supports the effectiveness of dual dopamine D2 and D3 antagonists or partial agonists in schizophrenia, depression, and bipolar mania. D2 antagonism is required for the antipsychotic effect, and D3 antagonism contributes to cognitive enhancement and reduced catalepsy.  Dual acting compounds should show higher activity to D3 than D2 (due to differential expression levels).  To this end, they apply their sequential docking protocol to identify potential points for fragment linking on the D3 crystal structure and D2 homology model.  These two targets have almost identical primary binding sites, but selectivity can be modulated through the secondary site.

In short, their in house fragment library consisted of 196 amine containing fragments for the primary site.  Second library of 266 fragments of cyclohexyl or piperidines.  Then, the first library was
docked to the apo receptor structures,then the docking poses were merged with the receptor, new grids were constructed including the merged ligands, and the second fragment library was docked to the partially occupied binding sites.  
Table 1.
As shown in Table 1, they synthesized three of their compounds and did generate potent and selective D3/D2 antagonists.  Linking is hard.   It still comes down to the right linker and all that entails.  Finding that right linker is made much easier by having structural data, as shown here.  This is a nice example of experimentally verifying in silico predictions. 

02 February 2015

Fragments vs DsbA: targeting bacterial virulence

The quest for new antibacterial agents can seem quixotic: no sooner have you found a killer molecule than the bugs have developed resistance to it. Evolution is hard to beat, particularly when it comes down to life or death. But what if you could lower the stakes? Many bacteria express virulence factors that are not essential for survival but are important for colonizing their host. Perhaps targeting these would be less prone to generating resistance.

Virulence factors often contain disulfide bonds, and the bacterial protein DsbA is essential for catalyzing their formation. In a paper published recently in Angew. Chem. Int. Ed., Begoña Heras (formerly University of Queensland), Jamie Simpson and Martin Scanlon (Monash University) and collaborators describe a fragment-based approach against the E. coli. version of this target. (See also here for Derek Lowe’s thoughts.)

The researchers started with an STD NMR screen of an 1132 fragment library from Maybridge, with compounds in pools of 3 to 5 (each at 0.3 mM). This yielded 171 hits, 37 of which showed appreciable chemical shift perturbations (CSPs) in a two-dimensional HSQC 15N-1H NMR assay. All of these were relatively weak, with none showing saturation at 1 mM fragment concentration, but they all appeared to be binding in a hydrophobic groove adjacent to the active site.

The 37 hits clustered into eight different structural subclasses, one of which – the phenylthiazoles – is described in detail. The Monash fragment library was designed with SAR-by-catalog in mind, and 22 commercial analogs were purchased and tested in the HSQC assay to assess the SAR. Several of the compounds were soaked into crystals of DsbA, in one case leading to a structure in which two fragments were bound stacked on top of each other in the hydrophobic groove. However, this binding mode was inconsistent with the NMR data, and indeed co-crystallography of the same fragment revealed a 1:1 complex with the protein, also in the hydrophobic groove. (As an aside, this is an interesting case of crystal soaking and co-crystallization giving different results; are readers aware of others?)

The crystal structure was used to inform fragment-growing, ultimately leading to molecules with dissociation constants around 0.2 mM as assessed by surface plasmon resonance and with similar IC50 values in a functional assay. One of these compounds was also tested against E. coli. DsbA is not needed for bacterial growth in rich media but is necessary for motility, and happily the assays showed just this – the compound did not affect growth but did inhibit cell motility.

Although the molecules are still too weak to answer the question of whether targeting DsbA will be a viable antibacterial strategy in vivo, this paper presents promising starting points, along with a wealth of data (including 78 pages of supporting information!) And if you want to learn more, Martin Scanlon is one of the organizers of the FBLD symposium at Pacifichem this December – so you can ask him questions in person in Honolulu!

28 January 2015

Get to Know Your Compounds

One of the first fragment screens I was ever involved with had RNA as the target (this is back when people did anti-bacterial research).  Because of that, I always try to write about targets, not proteins when referring to generic things we wish to find ligands for.  Nucleic acids have secondary and tertiary structure, just like proteins, and thus have ligandable pockets.  We have covered RNA as a target previously.  Well, we get to add another paper to the list.

In this paper from researchers at Goethe-Universität Frankfurt am Main present their results on HIV Tat-TAR.  This target was discussed over three years ago.  I am not particularly impressed with the compounds or the work (even though it included RNA NMR).  I was more impressed with comment made in the paper which hint at the kind of compound understanding we often cite as lacking from academic papers.

First, let's get to the guts of the paper.  They have been trying to identify ligands to this target for years, yielding nanomolar affinities but limited specificity.  Moving to fragments when all else fails, they decided to utilize very simple fragments: benzenes and amines, amidines, or guanidines able to be protonated at physiological pH.  Their fragments were screened in a fluorescent Tat-TAR-peptide assay.   Figure 1 shows the compounds tested.
Figure 1.  Compounds tested.  IC50s shown in parentheses. 
Cpds 1-6 were inactive, but cpd 7 looked promising...at first glance.  As the authors state:
"However, the IC50 value of this compound improves steadily when aqueous solutions are kept under air.  This effect was also found with other compounds, for example tetraaminoquinazoline 23.  The electron rich heterocycles in particular have the tendency to produce false positive results, presumably by forming positively charged oligomers."
I presume that they ran the assay and a few days later went back to re-test the compounds and saw anomolous results.  What really strikes me is that with a 7 mM IC50 upon retesting they saw a number sufficiently different (and that they trusted) to flag it.  They also note that such compounds must be carefully recrystallized and fresh powders ONLY used.   

The rest of the story is not nearly as interesting.  They performed 1H NMR titrations and 2D NOESY to confirm that these compounds are binding to the RNA.  They do some SAR and re-discover cpd 22, a known Tat-TAR inhibitor, that has already been patented.
So, what do we learn here? Understanding your fragments and their potential liabilities in your assay is crucial.