If you’d like to learn more about the topics under discussion at the Spring Meeting, the following literature may be of interest:
Patel, H.; Bodkin, M. J.; Chen, B.; Gillet, V. J., Knowledge-Based Approach to de Novo Design Using Reaction Vectors. J. Chem. Inf. Model. 2009, 49 (5), 1163-1184.
Soh, S.; Wei, Y.; Kowalczyk, B.; Gothard, C. M.; Baytekin, B.; Gothard, N.; Grzybowski, B. A., Estimating Chemical Reactivity and Cross-influence from Collective Chemical Knowledge. Chem. Sci. 2012, 3 (5), 1497-1502.
Berthold, M.; Cebron, N.; Dill, F.; Gabriel, T.; Kötter, T.; Meinl, T.; Ohl, P.; Sieb, C.; Thiel, K.; Wiswedel, B., KNIME: The Konstanz Information Miner. In Data Analysis, Machine Learning and Applications, Preisach, C.; Burkhardt, H.; Schmidt-Thieme, L.; Decker, R., Eds. Springer Berlin Heidelberg: 2008; pp 319-326
Herrgård, M. J. et mult al. (2008). A consensus yeast metabolic network obtained from a community approach to systems biology. Nature Biotechnol. 26, 1155-1160.
Thiele I et mult. al.: A community-driven global reconstruction of human metabolism. Nature Biotechnol 2013; 31:419-425.
Swainston N, Mendes P, Kell DB: An analysis of a ‘community-driven’ reconstruction of the human metabolic network. Metabolomics 2013; 9:757-764.
Kell DB (2013) Finding novel pharmaceuticals in the systems biology era using multiple effective drug targets, phenotypic screening, and knowledge of transporters: where drug discovery went wrong and how to fix it. FEBS J 280, 5957-5980.
Kell DB & Goodacre R (2014) Metabolomics and systems pharmacology: why and how to model the human metabolic network for drug discovery. Drug Disc Today, online.
Integrated Systems Approach Identifies Genetic Nodes and Networks in Late-Onset Alzheimer’s Disease, Cell, 153, Issue 3, , Pages 707–720 (2013) http://dx.doi.org/10.1016/j.cell.2013.03.030
The Psychiatric GWAS Consortium: Big Science Comes to Psychiatry – Neuron. Oct 21, 2010; 68(2): 182–186. http://dx.doi.org/10.1016%2Fj.neuron.2010.10.003
In Silico Target Predictions: Comparing Multiclass Naïve Bayes and Parzen-Rosenblatt Window and the Definition of a Benchmarking Dataset for Target Prediction. J Chem Inf Model 2013, 53, 1957–1966.
Cronin MTD, Madden JC, Richarz A-N (2012) The COSMOS Project: A Foundation for the Future of Computational Modelling of Repeat Dose Toxicity. http://alttox.org/ttrc/toxicity-tests/repeated-dose/way-forward/cronin/
Noel O’Boyle, NextMove Software
The concept of the Matched Molecular Pair (MMP), two molecules with the same scaffold but different R groups at the same position, has become very popular in recent years for rationalising trends in SAR. The success of Matched Molecular Pair Analysis (MMPA) is due to the fact that relative changes in property values are easier to predict than absolute values. Predictions based on MMPA work well for physicochemical properties as well as biological activities that correlate highly with such properties.
However, in general MMPA does not work well for predicting R groups that improve biological activity. The simple reason for this is that for one binding site environment changing group A to group B may increase activity while for another binding site environment it may decrease activity. While attempts have been made to address this problem, for example by focusing on MMPs from just the target of interest or with a particular atom environment, the underlying problem remains.
Enter the Matched Molecular Series (MMS), a concept introduced by Bajorath in 2011. This is simply a generalisation of the MMP concept to a series of any length, that is, N molecules with the same scaffold but different R groups at the same position. With MMPA, we are asking the question “Will changing B to C increase the activity?”; in contrast, if using MMS of length 3, we are asking “Will changing B to C increase the activity given that B is more active than A?” In other words, using longer series introduce more context, and this context represents a particular binding site environment.
Building on this idea in collaboration with AstraZeneca Mölndal, we have developed the “Matsy” algorithm (from “Matched Series”) which uses an existing source of activity data (e.g. the ChEMBLdb, or an internal database) to make predictions for what R group is likely increase activity given an observed activity order for measured R groups. Given a query matched series, the algorithm searches an activity database for all R groups that have been measured along with the query, and calculates the percentage of times each R group increased the activity beyond the most active R group in the query. The R groups with the highest percentages are presented as the most likely candidates to try next. For example, given an observed pIC50 order of ethyl > propyl > methyl, the top prediction is tert-butyl on the basis of 23 observations in ChEMBLdb of which 39% increased the activity.
In summary, using Matched Molecular Series you can overcome the limitations of MMPA for activity prediction by implicitly incorporating information on the binding site environment. The technique may be used as a way of guiding a medicinal chemistry programme, as a hypothesis generator, or simply a way to navigate existing SAR data.
For further details, please see our publication in J. Med. Chem. (http://dx.doi.org/10.1021/jm500022q) or the summary in a recent talk (http://www.slideshare.net/NextMoveSoftware/using-matched-molecular-series-as-a-predictive-tool-to-optimize-biological-activity).
Postdoctoral Researcher in Cheminformatics (3 years), Department of Environmental Chemistry at Eawag, the Swiss Federal Institute of Aquatic Science and Technology
The following meetings may be of interest to our readers:
CECAM Workshop on Entropy in Biomolecular Systems, Max F. Perutz Laboratories, Vienna, May 14-17 2014, http://www.cecam.org/workshop-0-1014.html
Cambridge Chemoinformatics Network Meeting, 28th May, Unilever Centre, University of Cambridge, http://www.c-inf.net/
RSC/MGMS Molecular Simulations & Visualisations, 7-9 May, Nottingham, http://rsc.li/fd169
20th EuroQSAR Symposium, St Petersburgh, Russia, 31st Aug -4th Sept, www.euroqsar2014.org
Full text of Rod’s bi-monthly newsletters, which include med chem literature reviews in addition to the more general papers included here, can be found on his website at http://rodporterconsultancy.com/newsletter/.
Miscellaneous developability – Toxcast, hepatoxicity, aldehyde oxidase
A report from Pfizer 1 describes initial results of an analysis of 52 compounds with preclinical and clinical data they submitted to the EPA Toxcast programme. The analysis identified particular in vitro assays (486 in total) run under the Toxcast programme of relevance to the Pfizer compound set. One important point was that the team identified differences in both chemical and biological space of pharmaceuticals compared to environmental chemicals which may make it difficult to interpret results for pharmaceuticals using in silico models developed with environmental chemicals. The analysis did identify novel interactions of the Pfizer compounds with multiple nuclear receptors in particular which had not previously been recognised.
Hepatoxicity is a significant problem in development of drugs. While there are some ways to check e.g. for reactive metabolites with glutathione trapping etc a more sophisticated alternative is now described 2. The approach uses microarray based toxicogenomics in an attempt to classify hepatoxic and non-hepatoxic using gene expression profiles determined in vitro from HepG2 cells. Use of 36 genes gave a 92% (training set) or 91% (validation set) accuracy. For cholestasis only 12 genes were required to detect 8 of 9 cholestatic compounds. Tentatively the group suggested that endoplasmic reticulum stress and the unfolded protein response re particularly relevant in predicting hepatoxicity although they do recommend that more compounds need to be evaluated
Finally in this developability miscellany an elegant strategy to identify aldehyde oxidase substrates 3 – see following discussion on rings in drug discovery. Aldehyde oxidase substrates are difficult to predict in silico and furthermore compounds show significant cross species differences in susceptibility. The approach exploits the proposed nucleophilic mechanism of AO and the likelihood that susceptibility to AO metabolism may be related to the susceptibility of a heterocycle to nucleophilic attack – particularly adjacent to ring nitrogen atoms. The team therefore used the observation of the predictable nature of alkylsulphinate derived radicals attacked heteroarenes to develop a chemical method for testing for introduction of a CF2H group to a molecule. Those molecules that generated an M+50 mass spec peak under a 2h incubation with Zn(SO2CF2H)2 were found to be aldehyde oxidase substrates while those that did not, were not substrates. False positives were also found as the introduction of the CF2H group was not as regioselective as AO nor was the chemical reagent as susceptible to steric interference as AO can be. None the less these data were a useful early warning. Furthermore the CF2H adducts formed were resistant to the effects of AO thus in one experiment you can identify your compounds liability towards AO and potentially fix it – ignoring any effects on biological activity just for now.
- F Shah and N Greene Chem. Res. Toxicol., 2014, 27, 86–98
- W. F. P. M. Van den Hof, et al Chem. Res. Toxicol., Article ASAP DOI: 10.1021/tx4004165 Publication Date (Web): January 28, 2014 Copyright © 2014, American Chemical Society
- F O’Hara, et al J. Med. Chem., 2014, 57, 1616–1620 DOI: 10.1021/jm4017976
Rings in drugs review diversity
In an analysis of rings, ring systems (complete ring or rings formed by removing all terminal and acyclic linking groups without breaking any ring bonds) and frameworks (all the ring systems but also includes ring systems that are linked by nonterminal acyclic groups) present in the FDA Orange Book for NMEs upto the end of 2012 a group from UCB drew several conclusions including;
- only 351 ring systems and 1197 frameworks have been used in current drugs.
- There are over 204 ring systems and 901 frameworks that have only been used once in a drug.
- On average six new ring systems enter drug space per year which has remained fairly constant over time.
- Each year, on average 72% of new drugs will comprise only those ring systems found in previously marketed drugs which sounds like a lot of cut and pasting!
- The overall count of 3, 4, and 5 respectively for O, N, and heteroatoms in a ring system has not particularly changed over time though I might have expected to see a trend to a slight increase in recent years with the focus on keeping lipophilicity down.
- Perhaps reflecting the privileged template idea ring systems cross therapeutic areas and target classes – If a ring system is reused, 62% of reuses will be for a different therapeutic area and 71% for a different target class.
I guess one problem with this analysis is it doesn’t necessarily reflect the number of new rings exemplified in, for example, patents. Are there, for example, developability liabilities of novel rings exemplified in patents that prevent progression of such compounds? Is there anything particularly significant about the rings and ring systems that do make it to market perhaps intrinsic metabolic stability or lack of P450 interactions? While developability issues might be a factor the concern does remain that chemists feel unable to commit to the time to develop novel rings under normal circumstances – a situation which, in my view, is unlikely to improve with so much outsourcing of chemistry.
This is an interesting read and reflects again the lack of new chemistries and perhaps the “unfashionable” nature of hetero(ali)cyclic chemistry and mirrors previous reports on the under exploitation of heterocycles. As a bit of a counterpoint is a review 2 on acylhydrazides as reagents for synthesis of O-, N- and S- containing heterocycles.
- R. D. Taylor, M. MacCoss and A. D. G. Lawson J. Med. Chem., Article ASAP DOI: 10.1021/jm4017625 Publication Date (Web): February 17, 2014 Copyright © 2014, American Chemical Society
- P. Majumdar et al Chem. Rev., Article ASAP DOI: 10.1021/cr300122t Publication Date (Web): February 07, 2014 Copyright © 2014, American Chemical Society
CNS delivery a break through?
CNS targeted biologicals should have enormous therapeutic potential but for the problem of delivery of adequate reagent. The idea of using the transferrin receptor to carry ligands into the CNS has been around for decades but never really caused much of a stir. Now, however, Roche/Genentech report 1 a monovalent fragment of a transferrin receptor antibody as a Brain Shuttle module to deliver a standard therapeutic antibody to the CNS which has caught the eye of various groups e.g 2, 3. A divalent antibody ends up being sorted into the lysosome with presumably subsequent destruction. Use of the Brain Shuttle modified Aβ antibody gave a 55 fold increase in target engagement relative to the non-modified antibody which translated into an increased reduction in amyloid. This all sounds very encouraging for CNS biotherapeutics but of course this is early days.
Intranasal delivery of a lactose analogue of endomorphin 1 via the olfactory epithelial pathway has been reported 4 with drug appearing in the olfactory bulb within 10 minutes of dosing with negligible drug appearing in either the blood or in other regions of the brain. The compound has previously been reported to show activity comparable to morphine when dosed either iv or orally – perhaps surprising in view of the tetrapeptide/linker/lactose structure a total of ~72 heavy atoms and 14 H-bond donors.
Disrupting the blood brain barrier has been seen as a way of getting drugs into the CNS for a while with approaches like osmotic shock and treatment with bradykinin antagonists. Another approach now reported 5 is the use of synthetic E-cadherin peptide HAV. HAV has been demonstrated to rapidly increase the permeability of the BBB following iv dosing to give 2-5 fold increase in permeability to a low molecular weight gadolinium marker a high molecular weight marker and a P-gp efflux transport substrate. Effects on the barrier were fully reversed after 60 min and were not attributable to changes in cerebral blood flow. Unfortunately there were no reports of improved therapeutic efficacy of any pharmacologically active agent. I must admit opening the BBB even temporarily does make me uneasy after all it is there for a purpose but I guess for intermittent dosing with for, an example, an oncology product needs must.
Finally a nice review of dendrimers for facilitating drug delivery to the CNS 6. Emphasis of the discussion is particularly on targeting brain tumours. Also discussed are toxicity of dendrimers, biodistribution and transport mechanisms. There have been an enormous number of articles on nanoparticle delivery especially in the oncology area but space, time and life are all too short to review all the recent reports.
- J. Niewoehner et al Neuron 2014, 81, 49-60
- R. D. Bell and M. D. Ehlers Neuron 2014, 81, 1-3
- Drug Discovery Today 9th Jan 2014
- C. D. Cros et al Bioorg. Med. Chem. Lett., 2014, 24, 1373-1375, 2014
- N. H. On et al. Mol. Pharmaceutics, Article ASAP DOI: 10.1021/mp400624v Publication Date (Web): February 19, 2014 Copyright © 2014, American Chemical Society
- L. Xu, et al ACS Chem. Neurosci., 2014, 5 , 2–13
P-gp – not just for the CNS
Removal of efflux liabilities from a series isn’t always easy and is relevant to CNS and non-CNS intracellular targets 1. A new predictive model may therefore be of some assistance built from 423 substrates, 399 non-substrates and 735 P-gp inhibitors. Non-substrates seemed to have lower molecular weights and higher solubility than substrates – frankly not very useful on its own but interesting was the discrimination of substrates and inhibitors. This suggested that the latter were more hydrophobic than substrates and further more substrates had an increased proportion of H-bond donors with particular spatial patterns. This is somewhat reminiscent of the work from Anna Seelig some years ago e.g. 2.
The potential for drug-drug interactions with P-gp substrates/inhibitors has been a source of some debate. New work 3 using PET indicates that this effect can be small as in the case of verapamil (substrate) and cyclosporine (inhibitor) However this work also suggests that these interactions are more likely to be a concern where fractional contribution of P-gp to CNS distribution is high – ft > 0.97 e.g. with nelfinavir. The authors predicted that cyclosporin (at clinically relevant blood concentration of 1.5 μM) would increase nelfinavir human brain concentrations by 236%. Presumably a similar argument might apply for intracellular targets.
Finally a nice development of serial sampling of CSF from rat cisterna magna to determine free brain concentrations 4. The technique avoids the problems of inter animal variance and saves on animal useage and was corroborated by comparing results with discrete CSF sampling and from whole brain. Of relevance to the previous discussion P-gp substrates that were assessed showed little increase in CSF concentrations in the presence or absence of the P-gp inhibitor elacradir.
- D. Li et al Mol. Pharmaceutics, Article ASAP DOI: 10.1021/mp400450m Publication Date (Web): February 18, 2014 Copyright © 2014, American Chemical Society
- A. Seelig and E. Landwojtowicz E. J. Pharm Sci 2000, 12, 31
- P. Hsiao and J.D. Unadkat Mol. Pharmaceutics, 2014, 11, 436–444
- T. Thanga Mariappan et al Mol. Pharmaceutics, 2014, 11, 477–485
More ligand efficiencies
Another review 1 on ligand efficiency metrics in drug discovery albeit with an impressive list of authors. Focus is particularly on various definitions of ligand and lipophilic ligand efficiencies alongside discussion of group efficiency and size independent ligand efficiency. There are some useful tabulations of changes in affinity required to maintain constant LE and LLE for different substituents. The authors analysed 59 optimisation programmes where LLE was used as a guide and highlighted that in the large majority of cases LLE was increased alongside increased target activity with most optimised compounds showing a reduction in cLogP – median increase in LLE of +1.96 with a p(activity) of +1.22. Interestingly the mean LE did not change during these optimisation programmes. Using a slightly different approach looking at the drug target level similar it was also concluded that increase in afifinity does not mean an increase in bulk physical properties. In a further analysis this time comparing ligand efficiencies of marketed drugs with other compounds binding to the target it is evident that for the majority of targets the marketed agent is amongst the most highly optimised with respect to LE and LLE. The notable exception is kinases where there tend to be large numbers of compounds in the literature with higher LE and LLE than the marketed exemplar. This may reflect the challenges of identifying compounds with appropriate selectivity profile for example. Finally looking at sets of compounds progressed to the clinic for the same target the authors point out that the compounds that have failed have been sitting in less optimal LLE/LE space than competitors that are still progressing. This really needs revisiting when clinical studies have completed.
All in all a useful read but the last word has to go to Hansch 2 and a quote from one of his (nigh on 30 year old) papers “Without convincing evidence to the contrary, drugs should be made as hydrophilic as possible without loss of efficacy”
- A. L. Hopkins et al Nature Reviews Drug Discovery 2014, 13, 105–121 doi:10.1038/nrd4163
- Hansch C., Bjorkroth J., Leo A., J. Pharm. Sci., 76, 663 (1987)
Druggable protein-protein interactions
From hot spots to hot segments PPI’s continue to attract increasing interest – specific examples later. A timely review 1 takes a look at what makes a PPI druggable and concludes that PPI’s with a dominant epitope (segment) is more likely to be druggable. Furthermore the authors argue that many PPI’s fall into this category and that peptides based on the interacting segments can form an accessible starting point for inhibitors. I certainly agree with that latter point to the extent of provision of tool compounds, however, how easy it is to then convert such peptides into (preferably oral) drugs is another matter either by peptide modification or by conversion to a peptidomimetic. Certainly one of the areas of medicinal chemistry we are not so good at addressing is rapidly generating good peptidomimetics.
Yet more convenient timing therefore a review by Hamilton et al 2 looks at α-helix mimetics shown to be present in about two thirds of protein-protein interactions. Early helix mimetics focused on presenting suitable side chains on one face i.e. i, i=4, i+7 only but many helix mediated PPI’s interact through two or three surfaces ie i, i+1, i+2, i+3, i+4 on the often amphoteric helix. Examples of helix mimetics are given but there are a number of downsides not least the atom efficiency is low with many heavy atoms to generate template rather than direct interaction. Not considered perhaps as much as I would have thought were the helix mimetics which have evolved from the work on P53/MDM2 inhibitors with four clinical candidates on the go examples of which are in Scheme 1. The latest of these is reported by Amgen 3 which evolved from targeted optimisation of the N-substituent giving a substantial increase in target activity, cellular potency and in PK profile. The sulphone side chain accesses a new so called glycine shelf pocket that other inhibitors do not seem to reach (PDB 4OAS). Using a rather different template derive from virtual screening and an ester prodrug strategy the simple indole acid template (prepared using an Ugi synthesis) YH239 has been characterised (PDB 3TJ2) as showing activity comparable to the Nutlins in patient derived AML samples 4. Just aking the cut for this newsletter a review 5 discussing different strategies to “drugging the P53 pathway” including MDM2 inhibitors that are attempting to activate or restore P53 action.
Finally while P53/MDM2 interaction have received intense scrutiny potent ligands to inhibit other PPI’s are also continuing to emerge. Not least inhibitors of the interaction between glucokinase and glucokinase regulatory protein that bind to the GKRP via a previously unrecognised but well-defined binding pocket distinct from the sugar binding region – see PDB 4MSU and 4LY9. From an HTS hit (1) target initial optimisation led to an increase in Alpha screen target activity of >100 fold and cell activity by about 50 fold with AMG1694 (2) although rat PK was poor 6. However, addressing metabolism by removing the morphline, stabilising the thiophene and methyl to trifluoromethyl conversion gave AMG3969 (3) with comparable in vitro profile but a dramatically improved rodent PK profile and a reduced heavy atom count 7. AMG3969 showed activity in a mouse model of diabetes with no evidence of hypoglycaemia.
- N. London et al. Curr. Opin. Cell Biology 2013, 17, 952
- M. K. Jayatunga et al Bioorg. Med. Chem. Lett., 2014, 24, 717
- D. Sun et al J. Med. Chem., Article ASAP, DOI: 10.1021/jm401753e Publication Date (Web): February 05, 2014 Copyright © 2014, American Chemical Society
- Y. Huang, et al ACS Chem. Biol., Article ASAP DOI: 10.1021/cb400728e Publication Date (Web): January 17, 2014 Copyright © 2014, American Chemical Society
- K. K. Hoe et al Nat. Rev. Drug. Disc. 2014, 13, 217
- K.S. Ashton, et al J. Med. Chem., 2014, 57, 309
- D. J. St. Jean, et al J. Med. Chem., 2014, 57, 325