Introducing Andy Smith, Discovery Services Scientist

The last time I worked for Cresset I ended up being outsourced to a customer – but it was all part of the service! I’m delighted to be back at Cresset Discovery Services to work on a new set of customer projects. Here’s a brief overview of my computational chemistry journey so far.

An early sniff at GPCRs

I was introduced to computational chemistry when I stayed on at Warwick for a research MSc. The project was an SAR analysis of bell pepper odorants predicting the structural and electronic features that are important for olfactory recognition. My sponsor, Ernest Pollack, was getting very excited about proteins called GPCRs, which had just been identified as the receptors responsible for smell, and could possibly be the largest collection of proteins yet found. Meanwhile, I moved on to a PhD using quantum mechanics to investigate the Heck reaction.

Working closely with synthetic chemists

In 2001 I joined Proteus (later Protherics, then Tularik), who had a history with Factor Xa and already had a licensing agreement with Lilly for LY517717 established. After the takeover by Tularik the focus moved onto kinases which were difficult for our in-house software to model accurately. It was here that, with help from Mick Knaggs, I moved back into ligand-based design. I started homology modeling to augment the range of targets available for the in-house structure-based design methods and to expand the range of target we could work on using ligand-based design tools.

A narrow focus on structure-based design

I joined Sterix/Ipsen in 2004. Sterix had three areas of research: tubulin, hydroxysteroid dehydrogenase and dual aromatase-sulfatase inhibitors. Frustratingly, our work on these projects was limited to structure-based design, due to the fact that the group had made a large investment in the GOLD docking package from CCDC.

Back to the full range of computational chemistry methods

Upon moving to Peakdale Molecular in 2006 I worked on a range of diverse and interesting projects using a wide range of computational chemistry methods, including structure-based, ligand-based, ADMET predictions, homology modeling, molecular dynamics, quantum mechanics, cheminformatics, bioinformatics, library design, diversity and similarity analysis. The diversity of the projects meant I quickly learned to adapt to customer priorities, selecting the most appropriate method for the best possible modeling solution, whilst maintaining value for money.

It was here that I first used Cresset software. Torch (then FieldAlign) and Forge (then FieldTemplater) provided intuitive and visually informative methods for analyzing molecules which were understandable to our customers. These applications offered very useful methods of treating molecules which were different from other approaches I’d used before; both tools gave a more insightful understanding of ligand similarities and a deeper understanding of the features driving activity across a wide range of chemotypes and targets.

My first move to Cresset

Redx were working closely with Cresset, but were unsure how much support they would need, so in 2013 I took on the joint role of computational chemist at Redx and application scientist at Cresset.

Redx had fully functional medicinal chemistry projects in anti-infectives, oncology and crop protection, with large repositories of data. The established computational methods from Cresset Discovery Services showed the potential of computational chemistry and was the reason for Redx wanting to bring computational chemistry in-house. I provided the computational chemistry component and established a close working relationship with the medicinal chemistry teams.

Spark, Cresset’s scaffold hopping and R-group exploration application was integral to the development of a patent busting portfolio. As the company matured, Spark was used for backup series generation and to prevent the same patent busting approach to be applied to our own patents.

As an application scientist at Cresset, I carried out a very preliminary investigation of protein fields to determine whether we could transfer the Cresset methodology from ligands to proteins. It’s very gratifying to see the evolution of this early work brought to market in Flare, which provides fresh insights into structure-based design by integrating cutting edge approaches in an accessible and flexible user interface.

Outsourced to a customer

I joined Redx full time when it became clear that they needed more computational support. Working closely with Cresset software support, synthetic chemists were trained on Forge, a powerful ligand-focused workbench for SAR and design, and were encouraged to undertake general computational chemistry project related tasks. This freed up my time to provide more in-depth investigations to several key problems on projects.

A significant success was a compound that was sold to LOXO Oncology for $40 million. The computational work for this project was originally outsourced to Cresset Discovery Services. The compound was developed in 5 years and is a testament to the quality science undertaken at Redx.

In fact, the Redx CEO at the time was so impressed with the contribution of Cresset Discovery Services that he wrote:

“Cresset is a valuable partner in our drug discovery programs. Their deep knowledge of computational chemistry and its application to drug discovery is enabling us to progress multiple projects across a wide range of target classes very quickly and cost-effectively.

“Cresset has consistently delivered insightful support to our drug discovery efforts. In our experience, their knowledge, responsiveness and collaborative approach have set them apart from their competitors.”

 Dr Neil Murray, CEO, Redx Pharma, UK

Back to Cresset Discovery Services

I was delighted when I heard that Cresset Discovery Services was looking for a computational chemist to join the team. I’m very pleased to be back to working on a wide range of challenges on varied customer projects and putting my skills to use in the real world of modeling.

Contact us for a free confidential discussion to see how Cresset Discovery Services can work alongside your chemists to solve problems, provide fresh ideas, remove roadblocks and add direction and insight to you project. If you’re based near Alderley Park and would like to meet for a confidential chat, get in touch as I’m based at the BioHub part-time.

Artificial Intelligence and machine learning: Where’s the intelligence?

“Of course we have Artificial Intelligence (AI), who doesn’t?” This was a throw-away line, during the closing remarks at a recent conference about future directions in medicinal chemistry. Currently it seems that every company inside drug discovery has an AI project or collaboration. But what is AI in drug discovery, what is it being used for, what are the challenges and what are the opportunities?

The distinction between AI and machine learning

Whilst the phrases ‘Artificial Intelligence’ (AI) and ‘Machine Learning’ (ML) are often used interchangeably, they are subtly different. A true AI method would have the ability to make decisions and propose new ideas from outside its knowledge base, whereas machine learning methods can only use the information from within their knowledge base. In fact, many AI systems are built on a foundation of one or more ML techniques – particularly neural networks. If you were to give the two methods human characteristics, you could say that AI is imaginative and might suggest a completely new series, whereas machine learning is intuitive – it could identify improvements within a compound series, but it would be limited to the transformations present within the training set.

Given that presently, most applications referred to as using ‘AI methods’ in drug discovery utilize modern machine learning methods, we will also utilize the shorthand of talking about “AI” to mean machine learning methods such as ‘deep learning’ which is a method of training very complex neural networks to find patterns in large volumes of complex data (so called ‘big data’). Most of these applications in drug discovery could equally utilize other machine learning methods, such as regular neural networks, support vector machines or random forest regression models as data sets are generally not that complex and not that large. However, there are areas where AI methods have real traction, for example for those working on genomics data, the separation in performance between traditional machine learning and modern AI methods starts to become tangible due to the large size and complex inter-relationships of the data sets involved.

Mining big data in drug discovery

The biggest promise of AI and machine learning for drug discovery is the capability of mining big data. AI methods such as deep learning have been deployed effectively in other industries, most notably in social media interactions.

Most established pharmaceutical companies have large repositories of data stored in large corporate databases built up over many years. This data tends to be noisy simply because of the volume, number of sources and the long time frame during which the data was collected. AI methods are tolerant and accepting of this type of data, providing the data sets are large enough, which makes them eminently suitable for extracting and analysing data from corporate databases. The ability to leverage this corporate knowledge to current targets has the potential to provide tremendous benefit, both in direction and speed of development, which translates to lower development costs.

Over recent years, a number of leading companies have formed strategic partnerships to investigate whether these methods can be transferred to drug discovery (Table 1). AI and machine learning methods are being deployed in a diverse range of areas, ranging from identifying compounds from existing libraries, target identification, target validation to new target discovery. There are also genomics companies who are looking to provide custom drug therapies based on an individual’s bio markers.

However, as with any emerging technology, there are very few case studies where these methods have successfully been applied, so at present these methods remain exciting, but generally unproven.

Table 1: Non-exhaustive list of companies who have formed strategic partnerships involving use of AI.

Company Partner Date Started Target Area
Roche1 GNS Healthcare 2018 Drug candidates
Pfizer2 IBM 2016 Drug candidates
Sanofi3 Berg 2017 Vaccines efficiency
Merck4 Numerate 2012 Drug candidates
Amgen5 GNS Healthcare 2018 Drug candidates
AstraZeneca6 Berg 2017 Drug candidates
Evotec6 Exscientia 2016 Drug candidates
Genentech8 GNS Healthcare 2017 Target validation
Takeda9 Numerate 2017 Target validation



Cloud Pharmaceuticals12




10 disease targets

Biological targets

Multiple targets


Prediction and understanding for smaller data sets

Big data applications are not the sole domain of AI and machine learning; smaller data sets can also be used to generate focused models which are suitable only for a single chemotype. These models can be used to progress a series toward the desired goal, usually removal of some unwanted toxicology, and there are numerous examples where this has been applied – both successfully and unsuccessfully.

Most computational chemists have, at some point, applied machine learning methods to a particularly intractable problem. However, it tends to be a second-tier resource, not because these methods are particularly difficult to employ, but because extracting understanding from these methods is often difficult. Whilst the predictions these models generate tend to be highly reliable, understanding the factors governing the predictability can be difficult or impossible depending on the properties used. If the properties used are chemical drug relevant properties (e.g., logP, PSA, pKa, etc.) then the underlying factors controlling the property under investigation maybe illuminated, but the predictability of the model tends to be lower. If, as often happens, all possible calculated molecular properties are used, then the model predictability tends to increase but the underlying factors controlling the models predictive mechanism can remain a mystery and the user gains no practical insight into the problem, just a method of prediction.

Therefore, there could be a lot of potential in analysing the AI systems currently in operation to confirm the validity of their predictions. It may be possible to use these methods in a way that makes it possible to glean more information about the interplay of factors that govern the predictability of the models, making them more useful to computational chemists.

AI and machine learning at Cresset Discovery Services

Cresset Discovery Services have successfully used machine learning methods when the data is amenable. Of course, producing a ‘black box’ predictive model is not generally enough for us or our customers. Instead, we look for a deeper understanding so a customer can make informed decisions in the context of the rationale behind the predictions made.

AI and machine learning methods are exciting new tools in a growing tool box of approaches which we can apply to bring value to customer projects. The real intelligence lies with the operator in being able to select the appropriate tool for the job. Another comment from the conference mentioned at the start of this blog post: “will AI techniques replace chemists in drug discovery? Probably not for the foreseeable future, but it is clear that chemists who use AI will replace chemists who don’t in the very near future.” Contact us for a free confidential discussion and to discover how we can apply intelligence to your drug discovery challenges.


Review of the XXV EFMC International Symposium on Medicinal Chemistry 2018

The September meeting of EFMC-IMCS was in the picturesque city of Ljubljana, nestled amongst the surrounding mountains, with its many bridges, hilltop castle and numerous riverside eateries. The good food, wines and beers alone would have made the visit worthwhile, but there was more than enough excellent science to keep us in the conference centre.

With three separate scientific tracks: Therapeutic Areas, Technologies and Chemical Biology, it was impossible for the 800 plus delegates to see everything (much like an ACS meeting), but the busy schedule meant that there was something for everybody.

A cracking start

Fittingly, the opening on Sunday included a fascinating talk by the Slovenian academic and Zika virus expert Prof. Tatjana Avsic-Zupanc (University of Ljubljana) who was the first to characterize the link between Zika infection and foetal abnormalities and to identify a causative molecular target: PrM protein mutant S139N.

Bayard Huck (Merck Biopharm) presented his interpretation of their strategy towards the evolution of the medicinal chemist, a technology and multi-discipline embracing ‘Versatilist’, the scientist equivalent of a ‘Swiss army knife’. Bayard went on to describe virtual screening (one of Cresset Discovery Services’ areas of expertize) as vital for increasing cost-effectiveness. He also described the current mania with all things AI as being at the peak of inflated expectations in the Gartner hype cycle, although he noted that useful applications are emerging from the hype. Bayard elaborated further on the nature of workspace and how the removal of ‘silos’ is essential for effective innovation to occur between scientists, which sounded remarkably familiar to me; I realised later that it was the open structure of the Cresset office I was thinking of.

Prof. Christa Müller’s (University of Bonn) talk about purines: P2X, P2Y, P1 and P0 related receptors and enzymes, and Prof. Peter Seeberger’s (Max Planck Institute of Colloids and Interfaces) talk about the development and use of a remarkable, state-of-the-art automated glycan synthesiser for synthetic vaccine discovery got the conference off to a cracking start.

First disclosures, GPCRs and other highlights

There were several themes running through the conference including: immuno-oncology, a real opportunity to tackle cancer, targeted degradation as a completely new therapeutic paradigm and the emergence of various AI strategies. Marwin Segler (Benevolent AI) showed how deep learning could be used for evaluation of millions of possible synthetic routes to a novel compound. First time disclosures are shown in table 1.

Table 1: 2D structures of first disclosures.​

Entry Description Structure
1 Risdiplam (RG7916); developed as a splicing modifier for the treatment of spinal muscular atrophy (SMA), from Hoffmann-La Roche and PTC therapeutics
2 Highly selective covalent BTK inhibitor LOU064 from Novartis  
3 Potent Novel Oral sCG Activtor from Bayer: Bay1101042
4 PDE9 selective inhibitor BI409306 by Boehringer Ingelheim for CNS disorders  


Risdiplam has the electrostatic hallmarks of an RNA/DNA nucleotide base interactor (a relatively electron deficient heterocycle which is then more suited to face on face interaction with nucleotide bases) in a similar manner to the fluoroquinolone antibiotics e.g., grepafloxacin (Figure 1).

Figure 1. Risdiplam structure (left) with electrostatic surface (red = positive and Blue = negative) and that of Grepafloxin (right).

GPCR talks that caught my eye are captured in table 2 and include: A MgluR2 ligand talk by Jose Cid (Jannsen), a GabbaB ligand talk by Sean Turner (AbbVie). Also, a very nice talk by Anne Valade (UCB) describing deuterium exchange as a method for identifying the allosteric ligand binding site for the dopamine D1 receptor. Finally, a first disclosure from Stefan Bäurle (Bayer) on their collaboration with Evotec.

Table 2: GPCR ligand structures.

Entry Description Hit Optimized lead
5 Metabotropic glutamate receptor 2 from Janssen
6 GabbaB positive allosteric modulators from AbbVie
7 Bradykinin B1 antagonist first disclosure Bay-840 from Bayer-Evotec

Other ligands captured from talks that were interesting are shown in table 3. Of particular interest was the superb talk by Allan Jordan, Cancer Research UK, which possibly wins the prize for the ugliest looking validated hit.

Table 3: Ligands from other diverse talks of note.

Entry Description Hit Optimized lead
8 David Tully from Novartis described their FXR agonist for NASH: Tropifexor
9 Anna Miggalautsch Graz University  Adipose trigliceride lipase – mouse active Atglistatin
10 Paul Fish from UCL Notum inhibitor for AD
11 Allan Jordan NMT1 inhibitor CRUK
12 Rainer Machauer disclosure on CNP-520 a new Bace2 inhibitor from Novartis
13 Juraj Velcicky Gamma Secretase-like membrane bound peptidase SPPL2A developed from Gamma Secretase SAR, for Endometriosis


Overall it was a successful, enjoyable and informative EFMC-ISMC symposium.

Modeling ‘big’: Applying the XED force field to biologics

Cresset is well known for powerful and accurate ligand-centric modeling, and Flare has established our methods for protein-ligand interactions. Work on GPCR modeling and viruses demonstrates the effectiveness and potential of Cresset technology for protein-protein interactions. Here I discuss the successes and challenges of modeling ‘big’ – applying Cresset’s XED force field to biologics.

Adventures in protein modeling: GPCRs

In 2014 Dr Andy Vinter, Cresset founder, reported on GPCR modeling exercises using the XED force field1, where ligand poses were exhaustively explored together with full complex minimizations to provide qualitative or quantitative analyses with binding estimates for agonist v antagonists. Although this was a huge modeling challenge, the approach provided fascinating new insights into GPCR behaviour that are in keeping with more recent literature. In particular, Brian Kobilka (joint winner of the 2012 Nobel Chemistry Prize) published a paper in 2016 showcasing the use of specific nanobody binding to the intracellular side of the GPCR to probe the long-range influence of ligands at the extracellular side2. He provided evidence supporting the hypothesis that GPCRs are likely partitioned between different states by differential stabilization of the full complexes in response to ligands. Our modeling findings concur in that the subtleties of these interactions extend beyond direct local binding interaction events and are propagated at distance across the full protein complex.

A matter of scale

Long-distance effects are not unusual in the realm of protein-protein interactions yet are beyond the scope of traditional molecular mechanics – from an accuracy point of view. From a sheer scaling point of view, the number of atoms involved means they are also beyond the scope of quantum mechanics. QMMM methods are also sometimes a poor compromise as these are discontinuous and focus on the local binding event.

Interestingly, this is where the XED force field has a nice sweet spot; accuracy approaching that of QM, but speed and the ability to map larger numbers of atoms >30,000, which is highly appropriate for the analysis of protein ligand and protein-protein systems. We can do this accurately and consistently through deployment of careful protein preparation and minimization workflows on protein systems.

Example: Influenza virus

The Centre for Pathogen Evolution at The University of Cambridge3 is involved in mapping data ultimately for the potential prediction of vaccine escape mutations of the influenza virus.

Hemagglutinin virus protein is the receptor that recognises mammalian cell surface glycans as an essential route to host cell entry. The ability of the virus to recognise sialic acid containing glycans is essential to this process and residues that contribute to its recognition represent those which are consequently difficult to mutate without compromising the virus. Antibodies which are directed to this site (Figure 1 left and middle) are less likely to suffer from viable virus mutations than others 4,5 (Figure 1 right).

Figure 1: Left: Influenza H3N2 hemagglutinin with the electrostatics of the core recognition element sialic acid from PDB 5VTQ. Middle: overlapping monoclonal antibody (blue tube) recognition site with electrostatics from key residues from PDB 2VIR. Right: a non-overlapping monoclonal antibody from PDB 5W42.

Optimizing biologics using 3D electrostatic shape and complementarity

In vogue, directed degradation mechanisms (PROTACS)/antibodies/vaccines, i.e., biologics, are example therapeutic paradigms which involve subclasses of these protein–protein interactions rather than the classical small molecule drug – protein target interactions. Modeling them is a significant challenge faced by many organizations charged with producing an array of diversely targeted therapeutics, because it is where a lot of what remains (the ‘higher hanging fruit’) happens to be.

The biologics industry may have slightly different criteria for cycling through an optimization, but ultimately similar schemes to those operating in the pharmaceutical industry still apply. There is an equivalent of traditional medicinal chemistry drug discovery workflows – involving SAR analysis, design, synthesis and test cycles. For antibodies, as for small molecules, target affinity, solubility, aggregation are key initial concerns. Mouse to human transformation is a uniquely biologics issue (unless we are talking in vivo models) as is the means of controlling SAR. For proteins it is all in the manipulation of the amino acid sequence, protein loop conformational preference, by single or multiple residue mutation. Incidentally, conventional sequence similarity metrics are not a useful measure of a residues relative potential for interaction with ligands or proteins in active sites (despite often being the tool of choice for analyzing protein data), as they were derived from natural mutation propensity and that consequence on maintenance of protein architecture.

Ultimately, the mechanism of target engagement, the molecular recognition event, is through electrostatic and shape complementarity and is fundamentally the same 3D phenomenon that applies to small molecules. Cresset scientists have an outstanding track record of working on electrostatic and shape complementarity and have successfully applied these skills to protein-protein interactions.

In the last 12 months, Cresset Discovery Services has completed work on viral vaccine modeling and biologics modeling which have proved highly useful for clients. We matched observed binding events with calculated binding enthalpy trends and predicting a-priori the observed pattern of protein binding or unwanted peptide binding suppression. This has been done using WT or mutant proteins that we have successfully taken through analysis, modeling/design and client testing cycles. As you would expect, client confidentiality prevents us disclosing further details, but contact us for a free confidential discussion.


  1. Applying the XED molecular mechanics force field to the binding mechanism of GPCRs
  2. Allosteric nanobodies reveal the dynamic range and diverse mechanisms of G-protein-coupled receptor activation, Kobilka et al, Naturevolume 535, pages448–452 (21 July 2016)
  4. Substitutions Near the Receptor Binding Site Determine Major Antigenic Change During Influenza Virus Evolution, David F. Burke, Derek J. Smith et al, Science 22 Nov 2013:
    342, Issue 6161, pp. 976-979
  5. Diversity of Functionally Permissive Sequences in the Receptor-Binding Site of Influenza Hemagglutinin, Nicholas C. Wu Jia Xie Tianqing Zheng, Corwin M. Nycholat, Geramie Grande, James C. Paulson Richard A. Lerner and Ian A. Wilson, Host & Microbe 21, 742–753, June 14, 2017

Parkinson’s Virtual Biotech secures further funding for novel gene transcription modulators project

Following excellent progress in this drug discovery project that we announced in March, I am delighted that our unique software, virtual screening capabilities and highly experienced team have contributed to Parkinson’s UK securing further funding for the novel gene transcription modulators collaboration we are working on along with Selcia.

We look forward to continuing to deliver molecular modeling support for the next phase of this important project.

Review of 8th RSC / SCI symposium on kinase inhibitor design: Towards new frontiers

As kinases have become a well-established class of therapeutic targets, it’s probably fitting that the theme of the RSC kinase symposium this year was relatively conservative, although a reasonably diverse selection of kinases (table 1) and therapeutic areas (table 2) were covered. It is remarkable that so many ATP pocket blocking inhibitors (i.e., most examples described this year) have translated to clinical candidates and drugs on the market, despite the fact that kinase biology, particularly the role and mechanism of kinase modulation by multiple protein complex formation as a prerequisite for signalling, is not very well understood.

Table 1: Kinase targets 2018 symposium.

Targets Number Kinase
Lipid kinases 4 PI2K, PI3K, PI4K, ATM
Serine Threonine 10 MPS1 (TTK), ERK, PKC, AMPK, CDK, LRRK2, MLK, TTBK1, PINK1, mTORC
Tyrosine 3 ALK, EGFR, TRK


Table 2: Kinase therapeutic areas 2018 symposium.

Therapy Number
Oncology 9
Diabetes 1
Transplant rejection 1
Tropical disease 1


The talk by Craig Malcolm (Promega UK Ltd) perfectly illustrated the lack of understanding of which processes involving kinases are actually operating in cells i.e., those which may, or may not, be therapeutically relevant and accessible. He described an interesting ‘different take’ on the standard kinase biochemical assay by using florescent kinase inhibitor probes as a technique to assay kinase activity in the ‘live’ cellular context.

Lipid kinases

The talks by Klaus Okkenhaug (University of Cambridge and Babraham Institute, UK), and Roger Williams (MRC Laboratory of Molecular Biology, UK), offered in depth exploration of the intricacies of the lipid kinase family and the effect of the wider molecular and cellular context of these kinases on the therapeutic outcome.

Klaus described how diverse this family is firstly with the PI3K delta inhibitor Idelalisib (1), which in combination therapy with biologic Rituximab for Chronic Lymphocytic Leukemia (CLL), serves to draw cells out of the lymph nodes so that they can be targeted by antibody treatment. The role of PI3K delta in immuno-oncology was evident in combination with CSF1 as this synergizes knocking out tumor cell protection. Finally, Leniosilib (2), is an anti-inflammatory agent through selective PI3K delta inhibition.

In his talk, Roger described the VPS34 system and how the nature of multi-protein architecture governs specific membrane interactions and thus activation of this complex assembly.

Figure 1: Left – VPS34 (human PI3K) with inhibitor (PDB code 4PH4); center – yeast VPS34 / VPS15 complex (PDB code 5KC2); right – the larger yeast complex (PDB code 5DFZ) showing the context of PI3K catalytic domain (in grey across the structures).

In addition, James Reuberson (UCB, UK) went on to describe a PI4K 3-beta inhibitor for immune suppression (3) and Emily Hannan talked about Genentec’s selective PI3K alpha inhibitor and degrader (4) for breast cancer. The latter molecule stimulates removal of PI3K, presumably by stabilizing a conformation that can be more readily ubiquitinated.

Kinase activators

Exceptions to the general rule of ‘ATP site binding inhibition’ were two talks describing the discovery and development of kinase ‘activators’ of AMPK (5) and then PINK1 (6). AMPK is an unusual kinase with an allosteric modulatory site above the ATP binding catalytic site. PINK-1 is also an unusual kinase which has extended loops around the catalytic domain responsible for binding and phosphorylating ubiquitin. PINK-1 interacts directly with an E3-ligase at the surface of mitochondria and modulates mitochondrial autophagy. Youcef Mehellou, (Cardiff School of Pharmacy and Pharmaceutical Sciences, UK) described the use of novel ATP mimetic prodrugs which activate a form of PINK-1 that is mutated in Parkinson’s disease.

Kinase inhibitors for the CNS

Unsurprisingly, most talks focused on oncology as the major topic. Another sub-theme was accessing the CNS with kinase inhibitors for brain tumors or with CNS penetrating ligands for psychiatry or neurodegenerative diseases. Compound 6  is an example of a CNS penetrating pro-drug. Most other examples focused on optimizing inhibitors either by  multiple parameter optimization of calculated properties, or using experimental measurements such as compound measurement in brain slice homogenate as a component of their KPU metric. These include: a TTBK1 inhibitor from Lundbeck (7) for Alzheimer’s, the ATM inhibitor from AZ (8) for Glioblastoma, and the pan TRK inhibitor from Chugai (9) for brain metastasis. Note the absence of the classic donor-acceptor-donor hinge interaction motif.

Figure 2: Lundbeck TTBK1 inhibitor (7) for Alzheimer’s disease (left – PDB code 4NFN) shows non-classical hinge interactions which are more evident using Cresset electrostatic field visualization (right) in Flare (cyan: negative potential; red: positive potential).



Overall the symposium was interesting and informative.

Contact us for a free confidential discussion to see how Cresset Discovery Services experts can advance your project.

Review of 255th American Chemical Society Meeting

As a first time ACS delegate and speaker, and first time visitor to New Orleans, these were pretty awesome firsts! And for different reasons both eye opening. The cacophony of night life and live music on Bourbon Street contrasted with the splendour of Royal Street and Decatur Street architecture and many excellent cafés and restaurants. This, together with the hefty pace of an almost overwhelming barrage of scientific presentations, posters and exhibitors, was both exhilarating and exhausting.

Relaxing in my hotel room, as the sun glittered off the wide Mississippi below, I could reflect on the week’s events. Here are some of the highlights for me.

Cresset science

We presented in the CINF fragrances, food and cheminformatics session, work we have done to classify odorants using shape and electrostatics generated using the XED force field 1. The less intuitive relationships between structure and olfaction, which are often not correctly attributable using 2D similarity techniques, are adequately described by a 3D field and shape treatment.

In COMP, our poster on the addition of user defined pharmacophoric features to Blaze 2 (giving an additional way for users to influence the results beyond pure electrostatics and shape) received a lot of interest – particularly in the pros and cons of doing so. These new features were released in Blaze V10.3.

Fragrances, Food & Cheminformatics

One key issue in flavour & fragrance science is the absence of a really robust biological readout for quantifying odour perception. Jack Bicker (IFF) 3 and Joel Mainland (Monell Chemical Senses Center) 4 both discussed the ability to predict the intensity of odorants, and to reconstruct something akin to a full dose response curve, by manipulating ‘odour perception data’ collated from human panelists. Jack covered a range of predictive modelling techniques used for analysing both property and structure data for odorants. Joel Mainland’s remaining time focused on the results of the ‘Dream challenge’ which provided details of sensory data on a series of diverse odorants and invited ‘participants’ to predict the variety of odorant properties from structure alone.  Surprisingly, this appears to have been most successful via recursive partitioning techniques.

GPCRs revival

Over the whole ACS there was a revival of interest in GPCRs as targets including Merck presenting on CBR1 5, CCR2 6, GPR40 7 8. All very much on the back of advances in obtaining x-ray crystallography for this important family of proteins.

Oliceridine as an agonist of the μ-Opioid receptor, a brand new drug for quite an established target, was presented by Trevena 9. One of the key issues with the use of agonists of this receptor (e.g., morphine), which behave pharmacologically as analgesics, is an associated respiratory depression effect. Trevena developed their medicine specifically to avoid this side effect and were able to do so by optimizing the drug specifically for the beta arrestin signalling pathway.

Figure 1: Oliceridine in the context of μ-Opioid receptor PDB: 5c1m using Forge to align it to the original bound ligand from this PDB structure.
Similarly, this notion resurfaced during Merck 5 and Lilly 6 talks on the effect of glucose lowering for diabetes via engagement with the GPR40 receptor. Again, biased signalling was associated with achieving a beneficial outcome. Lilly described the observed relationship between off rate and signalling bias – suggesting that increasing ligand residency time was associated with an increased engagement of the beta arrestin pathway. However, in this case the relationship between SAR and the pharmacology is complicated further by the observation of two distinct allosteric binding locations for the GPCR ligands.

Figure 2: Merck’s two allosteric ligands bound into the GPR40 receptor PDB: 5tzy with the ligands shown in space fill (top structure magenta and bottom green).
Thus, the portfolio of allosteric binding sites for GPCRs is duly expanding (numbers in parenthesis); with B2AR (2) GPR40 (2), CCR9 (1), ACM1 (1) all showing allosteric sites including a number where protein lipid interfacial contacts are made with the ligands.

New targets

New targets (to me at least) described in the various medicinal chemistry talks include, to mention a few: APOBEC3 10, VMAT2 11, HPGDS 12, Ketohexokinase 13, SH2P 14 and CRB 15. Also, in particular, the DUBS and their corresponding ligase counterparts, which together control ubiquitin mediated signalling, received lots of attention. For instance, a fascinating talk from Michael Clague at Liverpool University’s Cellular and Molecular Physiology Department 16, inferred that USP30 has a potential role in neurodegenerative disease. USP30 is involved in regulating mitochondrial autophagy via modulation of the extent of ubiquitin labelling of external membrane proteins on mitochondria.

First time disclosures I noted were:


AstraZeneca’s AZD0364 as a potent, selective orally dosed ERK inhibitor for anti-cancer therapy 17:

GDC-0927; an estrogen receptor degrader from Genentec for breast cancer 18:

Two BET inhibitors from Incyte INCB057643 and INCB054329, showing anti-proliferative activities against hematologic malignancies and solid tumors in preclinical models, currently entered into human clinical trials 19:

An RSV (Respiratory Syncytial virus) fusion protein inhibitor RV521 from Reviral 20:

VNRX-5133, a broad spectrum beta-lactamase inhibitor that is active at the standard enzymes including metallo-enzymes NDM and VIM associated with ‘super-bugs’, from Venorx 21 ; showing the wide scope of the medicinal chemistry efforts against a variety of disease targets for both human and animal health and made for a very interesting and informative week:


  1. CINF 24: Using molecular fields to understand molecular determinants of the olfaction process
  2. COMP 192: Adding pharmacophores to shape and electrostatics – too much of a good thing?
  3. CINF 26: QSAR/QSPR models to support fragrance ingredient molecular design
  4. CINF 23: Predicting human olfactory perception from molecular structure
  5. COMP 413: Exploring the structural features of peripherally-restricted CB1 receptor antagonists and inverse agonists
  6. COMP 366: Use of modeling and crystallography to understand CCR2 antagonist SAR and receptor binding
  7. COMP 365: Targeting GPR40 for diabetes: A multipronged approach
  8. MEDI 307: Advances in understanding the relationship of pharmacology to GPCR structure and function: A GPR40 journey from the lab to diabetic patients
  9. MEDI 226: Discovery of oliceridine (TRV130), a novel G protein biased ligand at the µ-opioid receptor, for the management of moderate to severe acute pain
  10. MEDI 258: Discovery of small molecule and nucleic acid inhibitors of APOBEC3 deaminases
  11. MEDI 280: Discovery and characterization of ingrezza (valbenazine): A VMAT2 inhibitor approved for the treatment of tardive dyskinesia
  12. MEDI 274: Orally-active inhibitors of H-PGDS for the treatment of asthma, allergic rhinitis and chronic obstructive pulmonary disease
  13. MEDI 275: Optimization of the chemical matter and synthesis leading to a ketohexokinase inhibitor clinical candidate
  14. MEDI 277: RMC-4550: An allosteric inhibitor optimized for in vivo studies of SHP2
  15. MEDI 279: Rapid development of potent and selective CBP inhibitors: The impact of a tetrahydroquinoline LPF binder
  16. MEDI 233: Target selection and small molecule development in the DUB space
  17. MEDI 293: Discovery of AZD0364, a potent and selective oral inhibitor of ERK1/2 that is efficacious in both monotherapy and combination therapy in models of NSCLC
  18. MEDI 294: GDC-0927, a selective estrogen receptor degrader and full antagonist for ER+ breast cancer
  19. MEDI 306: Invention of INCB054329 and INCB057643, two potent and selective BET inhibitors in clinical trials for oncology
  20. MEDI 308: Design, identification and clinical progression of RV521, an inhibitor of respiratory syncytial virus fusion
  21. MEDI 309: Discovery of VNRX-5133: A broad-spectrum serine- and metallo-beta-lactamase inhibitor (BLI) for carbapenem-resistant bacterial infections (“superbugs”)

Resurrection of the covalent inhibitor?

There has been a revival of interest in covalent inhibitors in recent years. This has culminated in a growing list of diverse proteins that have been successfully targeted, that are outside the well-trodden protease area.1-4

An older non-protease example is Clopidogrel. This P2Y12 inhibitor requires activation in vivo (through oxidative metabolism) to generate the electrophile intermediate (sulfenic acid) that forms a covalent disulphide bridge to a cysteine residue in the target GPCR. Unravelling the true mechanism of this compound led to second generation inhibitors which further exploit the covalent mode of action5.

Potential benefits

One more recent, and well documented, non-protease example is in the field of oncology, with the drug Ibrutinib. The covalent binding of the inhibitor to the target kinase BTK demonstrates multiple benefits:

  • It removes competition with the endogenous ligand, as is true for other covalent inhibitors, but in this case is more significant as concentration of competing cellular ATP is in the millimolar range.
  • High efficacy, in contrast to some conventional inhibitors, gained specifically by using this covalent inhibitor design.

Figure 1: Ibrutinb in BTK PDB: 5P9J showing Cys-481 spheres near with the covalent linkage visualized using Flare a new structure-based design application from Cresset.6
From a safety point of view, a method of achieving high efficiency/potency of any drug should translate into a lower dose that is ideally below any potential toxicity threshold (very few drugs are given at =< 10mg per day!). Similarly, prolonged duration of action, a consequence of the covalent inhibition, can result in less frequent dosage regimes that could be very beneficial – depending on the exact pharmacology.

So why have drug discovery companies not adopted this approach more broadly?

Mainly, because this is not a panacea – some proteins are more quickly processed, turned over or are mutated. The BTK case itself is an unfortunate example of the latter; the resistance mutation C481S produces a protein that cannot undergo the reaction with the inhibitor and ultimately limits its utility in patients. What may have more generally hampered covalent inhibition strategies in the past, observed particularly in the field of protease inhibitor design, was the high electrophilicity and poor selectivity. The very justifiable fear of off-target activity and threat of unresolvable toxicity issues, has contributed to the adoption of a dogma which still significantly disfavors the approach across the industry. The lack of a wider uptake of the approach since then has been exacerbated, perhaps, by the industrialization of experimental design and/or the lack of attention to the kinetic behavior of these systems. One size doesn’t really fit all!

Computational chemistry is critical for modern covalent inhibitor design strategies

An appreciation of the kinetics of molecular recognition and its impact on drug efficacy (Figure 2) has had a resurgence; together with the introduction of biophysical techniques, such as Surface Plasmon Resonance (SPR), which allow the routine resolution of otherwise painfully elusive kinetic parameters such as kon/koff.

Figure 2: Schematic of the covalent inhibition process (ki/Ki equivalent to classical Michaelis Menten enzymic parameters kcat/Km), transition state energy diagram for the covalent transformation.

The observed rate of inhibition for a covalent inhibitor is effectively dependent on the ratio of the rate constant for the chemical transformation of the inhibitor (ki), relative to the rate of unbinding (koff). We can introduce a level of catalysis in this system if we design appropriate geometries that lower the energy of (i.e. favor) the E : I complex7. If we are talking about hand-crafting optimum molecular geometries, then we are usually thinking about utilising computational chemistry as a technique.

Advances in the covalent inhibition area have clearly arisen from a systematic investigation into the modulation of the intrinsic reactivity of these electrophilic ligands. Covalent inhibitor design, in its more current form, now reduces reliance on the Ki, and drives optimization through binding (i.e., reducing koff) rather than reactivity. This more thoughtful approach is an essential step towards maximizing selectivity and minimizing off target activity. In parallel with the ‘much needed’ experimental investigations, a range of computational techniques are required for modern covalent inhibitor design strategies. These include:

Ligand-based design

  • Virtual calculations defining reactivity and ranking of electrophiles.

Structure-based design

  • Optimizing the molecular recognition toward the target enzyme by careful design of the ligand functionality through steric and electrostatic control.
  • Optimizing placement of an appropriate electrophile on the inhibitor in relation to an accessible, non-conserved nucleophile on the protein. Selectivity of the covalent bonding step is governed by the choice of a suitable electrophile and proximity to the nucleophile. Optimizing the reactivity of the electrophile for a covalent inhibitor requires a balance between achieving a sufficient high rate of reaction with the selected nucleophile on the target protein as well as minimal rate of reaction with off target nucleophiles.

We anticipate that this approach will prove increasingly useful, particularly for otherwise intractable biological targets involving large and complex protein-protein contacts.

Add direction and insight to your projects

Cresset Discovery Services’ experienced scientists work alongside your chemists to solve problems, provide fresh ideas, remove roadblocks and add direction and insight to your projects.8 They offer expert molecular design capabilities and use cutting-edge technologies to provide support for numerous client projects, across a range of disciplines.

Free confidential discussion

To find out how Cresset Discovery Services can help you reach your next milestone faster and more cost effectively, contact us for a free confidential discussion.

References and links

  1. Bauer R.A., Covalent inhibitors in drug discovery: from accidental discoveries to avoided liabilities and designed therapies, Drug Discov. Today, 2015, 20 (9), 1061-1073.
  2. Baillies T.A., Targeted Covalent Inhibitors for Drug Design, Angew. Chem. Int. Ed., 2016, 55, 13408-13421.
  3. De Cesco S., Kurian J. Dufresne C., Mittermaier A.K., Moitessier N., Covalent inhibitors design and discovery, Eur. J. Med. Chem., 2017, 138, 96-114.
  4. Lagoutte R., Patouret R., Winssinger N., Covalent inhibitors: an opportunity for rational target selectivity, Curr. Opin. Chem. Biol., 2017, 39, 54-63.
  5. Exploring synthetically accessible alternatives to P2Y12 antagonists using electrostatics and shape.
  7. See ‘Hammond postulate‘ for an exothermic reaction.

Computational approaches to ion channel drug discovery

In October, I presented Computational approaches to ion channel drug discovery at the ‘USA Integrated Drug Discovery Spotlight Event: Overcoming Challenges in Ion Channel Drug Discovery and Safety’ hosted by Concept Life Sciences and Metrion Biosciences.


Structure-based molecular modeling approaches are probably the most acceptable form of computationally driven rational design deemed useful by the medicinal chemistry community for drug discovery. For unliganded projects, invariably, medicinal chemistry is a painstakingly slow process.

Ion channels are inherently difficult protein targets to tackle as very few have been crystalized. Strangely, it seems that the greater the knowledge we accumulate – the deeper the complexity.

Alternative approaches e.g., chemogenomics and good ligand centric approaches, are vital to squeeze out all valuable insights from this ‘costly’ SAR data.

Cresset has a reputation for producing powerful software for ligand-centric molecular modeling. Analyses involving molecular 3D shape and accurate electrostatics, via Cresset’s proprietary multipole force field (XED), allow fast and informative interrogation of SAR in the absence of structure data.

I describe the current diversity of the ion channel target space and highlight some of the issues faced. Flare expands Cresset’s product portfolio into structure-based molecular design, bringing our electrostatic workflows into the protein realm.  Electrostatics are fundamental to ion channel function. These techniques are highly likely to prove very useful in the ion channel drug discovery arena, as more appropriate techniques for providing structure data become available e.g., CryoEM.

I make the case for information based strategies and ligand centric techniques, particularly computational workflows used by Cresset Discovery Services for customer projects, which pit the efficiency of rationality against blind trial and error.

Flare enhances the Cresset Discovery Services toolbox

One of the advantages of working for Cresset Discovery Service is that I have access to software that is under development. I’ve therefore been working successfully with Flare, our newly-released structure-based modeling application, on client discovery projects for some time and I’m delighted that it is now available to a wider market.

Processing structures to extract discovery information

So, what does Flare offer Cresset Discovery Services that it didn’t have before? Forge is the comprehensive workbench for ligand-centric workflows and I use it as the main workhouse for ligand work. However, more and more drug discovery projects are now structure enabled.

In fact, the increasingly routine use of X-rays and Cryo-EM means that projects generally have access to lots of structures, not just one. Processing these in order to extract useful information is often a key step for us on client projects. For this structure enabled work, Flare has become my Forge equivalent – my main workbench for structure-based work.

Using Flare

The Flare GUI is intuitive and easy to navigate, yet already has significant, powerful and unique functionality. From a practical point of view, Flare is stable on loading 20 or more X-ray structures, and doesn’t slow down (unlike some other tools) or glitch. These structures can be automatically aligned by sequence (Cobalt) and superimposed at the click of button.

Protein preparation is automated. Water analysis via 3D-RISM allows exploration of key water molecules around ligands and binding pockets. Of course, Cresset’s own electrostatic fields can now be visualized on selected protein surfaces, in addition to ligands.

Lead Finder is a Flare plug-in that gives the capability to dock ligands and ultimately conduct high throughput structure-based virtual screening as a complementary addition to Cresset’s state of the art ligand centric-virtual screening service using Blaze.

Zika virus protease PDB : 5H6V and a covalent inhibitor. Protein positive fields (left), negative fields (middle) and ligand fields (right).

Flare is focused on ligand design

The addition of Flare to the Cresset Discovery Services toolbox enhances our client offering with a new degree of confidence for structure-based work. Flare has been developed based on input from scientists in industry and academia so that it is relevant to real discovery workflows. My own work on client projects also contributed to the design of Flare, and I know from personal experience that it is ideally suited to the needs of computational chemists.

The upshot of this is that Flare has been created with ligand design as a central task. This means that even the most computationally intensive tasks are tailored to the effect that they have on the molecules that you make. This really is the great power of computational chemistry; informing and enhancing the discovery of the molecules that matter to your project, using the best computational methods at our disposal. Flare greatly eases that task and is a great addition to our toolbox.

A single force field across ligands and proteins

However, what really puts Flare apart from other structure-based methods is the seamless use of the XED forcefield between ligands and proteins. This means that I can make calculations on structure-ligand interactions based on a continuous force field, and make meaningful comparisons at all stages of work.

Ever since I joined Cresset in 2012, customers have been asking me, ‘Why not apply the XED force field to proteins?’ I’m delighted that this question has finally been answered with the release of Flare.

Contact us for a free confidential discussion about how cutting edge ligand and structure-based methods can transform your discovery.