Medicinal chemistry meets systems biology

1
Medicinal chemistry meets systems biology
John Harris, cjh Consultants (Founder and
consultant to BioFocus)
Cutting Edge Approaches to Drug Design MGMS,
March 2009 School of Oriental and African
Studies, University of London
2
Why should drug discoverers bother about
biological networks?

• nearly all drugs can hit more than one effector
  target in an organism
• not all non-target effectors are off-targets,
  metabolic systems or transporters
• accumulated genomic/proteomic/analytical
  pharmacological knowledge confirm that several
  highly efficacious drugs exert their overall
  therapeutic effect through a network of effectors
• the output of the network determines the drug
  profile (i.e. its good points and its bad points)

3
How should they respond to the challenges of
biological networks?

• 1970-1990 clinical success driven by
  selectivity for single targets (e.g. h2
  antagonists, AII inhibitors). Medchem is driven
  by isolated enzyme assays or analytical
  pharmacology.
• 1990-2000 as therapeutic targets become more
  challenging, high-throughput screening, fed by
  massively combinatorial chemistry, drives
  expectations upwards BUT the same technology
  demands assay systems even less related to the
  constituted organism!
• 2000- 2005 unmet expectations drive a much
  more focused approach to screening but compounds
  are still, essentially, optimised against single
  reductionist assays.
• 2005- present increasing realisation that
  reductionist assays do not predict cell network
  responses primary cell screening begins to gain
  ground.

4
Many clues along the way..

• most of the clinically effective antipsychotics
  require polypharmacological mechanisms
  (clozapine, a broad-spectrum biogenic amine
  ligand, is as effective as 5HT2a selective
  atypical antipsychotics such as olanzapine,
  ziprasidone, etc. (see Roth et al.,
  2004NatureRevDrugDiscovery353)

• in anti-infective therapy, polypharmacology is
  common, e.g. Wellcomes Septrin (trimethoprim and
  sulfamethoxazole hitting the bacterial
  network) or various HIV therapies (NNRTIs and
  protease inhibitors)

• more recently, one of the earliest
  clinically-successful anticancer kinase
  inhibitors, Sutent, has been shown to be one of
  the least selective across the kinome

5
(Sauer et al., Science (2007), 316, 550) "The
reductionist approach has successfully identified
most of the components and many of the
interactions but, unfortunately, offers no
convincing concepts or methods to understand how
system properties emerge...the pluralism of
causes and effects in biological networks is
better addressed by observing, through
quantitative measures, multiple components
simultaneously and by rigorous data integration
with mathematical models"
Systems Biology and Network Pharmacology are now
very well established BIOLOGICAL activities in
academia and, increasingly, in pharma and
biotech. They are driven by major technology
advances in high-content cell screening, cellular
disease modelling and data handling/knowledge
extraction.
Whither systems medchem?
6
How should the medicinal chemist respond?

• Historically screen in a black box
  empirical SAR but high relevance and guaranteed
  efficacy
• Contemporary screen target in isolation
  precision SAR but relevance and efficacy
  unclear
• The compromise take secondary screening into
  the cellular context (still much scepticism about
  primary cellular screening!) really depends on
  the degree to which the cell assays reproduce the
  target disease

• So how DO we blend the efficacy lessons of the
  past, underpinned by network pharmacology
  evidence, with modern screening and secondary
  assay technologies?
• How much must we change our mindset? After all,
  we optimise activity and ADME/PK more or less in
  parallel these days is an extra parallel target
  or two a quantum leap?

7
Kinases show the way forward?

• Clinically effective first generation oncology
  drugs (e.g. Sutent, Sorafenib) act at
  several/multiple target kinases and mutants
• These earlier multiple kinase inhibitors (MKIs)
  were discovered serendipitously (see
  2006NatureReviewsDrugDisc835)
• How do we discover and design MKIs rationally?
  (see 2010JMC1413)
• The challenges
• Multiple target discovery theoretical and
  analytical
• Lead discovery cross-screening fragment
  re-assembly chemoinformatics
• Lead optimisation
• balance of activities into the nearly-unknown
• balance of physicochemical properties
• balance of off-target activities

8
It can be done!
Lapatanib designed to hit EGFR and ErbB2 in
order to cover a wider range of tumour types
(see 2005Drugs of the Future1225)
9
Target Discovery Approaches

• In silico
• predict therapeutically useful combination of
  targets by network modelling and simulation
• correlate with known drug profiles, protein
  interaction fingerprints, biomarker data
• key input from broad chemogenomic databases
  which correlate high-quality assay data and in
  vivo data (pre-clinical and clinical) with
  specific targets
• In vitro
• Isolated enzyme profiling is arguably too
  reductionist at best can only point to possible
  targets or pathways
• cell lysate fishing using ligand probes is a
  better indication especially if studying
  affinities and response time-course (e.g.
  Kinaxos KinAffinity, Cellzomes BioBeadsTM
• High-content screening in cellular disease
  models, tracking networks, not just specific
  targets
• Counter-screening using characterised probes

10
Fesik et al. 2006Oncogene1340 Akt-co-operating
kinases

• A-443654 was counterscreened against 768-O cells
  transfected with a kinome-wide (443 kinases 64
  orphans) siRNA library
• Akt-dependent apoptosis and blockade of critical
  Akt signalling pathway nodes were both sensitised
  by siRNAs encoding CK3g1 and IMPK (inositol
  polyphosphate multikinase)

11
Lead Discovery Approaches

• In general, diversity screening against multiple
  targets may be even less cost-effective that
  against single targets, and key intra-family SAR
  is unlikely to be revealed. Pre-filtering based
  on overlapping pharmacophores a better bet
• A rational approach to MKIs is possible
• Feasibility assessment
• b) Focused screening library and fragment
  cross-screening

12
Ligand SAR and cross-family common site sequences
Recent evidence supports the view that, within
protein families that have a common site of
action, similar ligands tend to bind to similar
family members (see Bamborough, 2008JMC7898
Vieth, 2005DDT839) BioFocus has developed a
simple roadmap based on the common geometry of
the kinase ATP site (activated state) which
enables quick assessment of multitarget SAR
crossover feasibility
13
Feasibility assessment A dual inhibitor of
LimK1 LimK2?
14
SoftFocus library screening gave 3 scaffolds but
only two subsite-selective fragment classes
(based on homology model docking)
15
Best dual activity for these U-shaped compounds
was 12-fold (in either direction) Subsequent
report showed nanomolar dual (equipotent)
inhibition for a series of linear compounds of
general structure
16
Focused library cross-screening

• Despite close sequences, SAR relationships for
  similar kinases remain unpredictable, especially
  where overt or more subtle differences in binding
  mode occur
• Many organisations have built up vast datasets
  of cross-screened kinase inhibitors which can be
  mined for MKI leads
• IPR break-out is more of a problem with this
  strategy though

17
Fragment screening and cross-screening

• Increasingly popular approach well suited to MKI
  design
• Fragments sample greater chemical space and
  allow identification of preferred monomers and/or
  monomers which may not be picked up when
  screening a more decorated system
• Fragments have no pre-determined second
  vectors and are able to probe sub-sites more
  extensively
• Compounds can be grown from common scaffolds
  or pre-determined privileged sub-site fragments
• Larger fragments can be screened in biochemical
  assays

18
ThemePair Fragment Libraries

• Small fragments
• 1 component system
• Require affinity-based techniques (HTX-ray, NMR,
  SPR)

• ThemePair Fragment library compounds
• 2 component systems
• lead-like compounds, good ligand efficiency,
  solubility HC-biochemical screening or affinity
  methods

• Traditional focused library compounds
• 3 component systems
• more likely to give potency in biochemical
  screens
• less likelihood/compound of multitarget SAR

19
Addressing multi-targeting in a rational way
using designed fragment libraries such as
ThemePair Fragments
Illustrative simplistic scenario
Primary kinase target Scaffold X, sidechains a-g
Secondary kinase target Scaffold X, sidechains
e-h
Exclusion kinase target Scaffold X, Sidechains
a,c,j
Therefore, profitable SAR area for selective
multi-targeted inhibitor is scaffold X combined
with sidechains e,f and g
20
In reality, likely that similar scaffolds will
show similar SAR at the themepair fragment level

• Favoured area of space for required hit profile
• Can provide a menu of scaffold and
  side-chain/monomer types

21
An example BioFocus library TPF11

• large library (ca. gt700 compounds) based on two
  cores and 9 scaffolds extensively elaborated at a
  single position so that the scaffold becomes the
  effective second variable. These compounds are
  not reported in SciFinder or by commercial
  supplier

22

• Lead optimisation where (medchem) going gets
  tough!
• Balance of activities into the nearly-unknown
  until more data are available from network
  biomarker and enzyme-occupancy studies, balanced
  potency is the best guess very high
  multipotency may well not be required
• Balance of physicochemical properties tricky
  for MKIs where structural additivity tends to
  correlate with selectivity however, deliberate
  choice of overlapping pharmacophores helps
  non-oncology applications are more challenging
• Balance of off-target activities this issue is
  no different in principle to that for so-called
  selective kinase inhibitors, of which there are
  not many. Isolated enzyme assays are, at best, an
  approximate guide to undesirable intra-family
  activities. Monitoring cellular target/s activity
  against in vitro and in vivo toxicity readouts
  are essential in lead optimisation.

23
Facilitating parallel lead optimisation

• Parallel optimisation is the ideal
• This is the area of greatest current medchem
  caution! Lead optimisation against more than one
  non-ADMET/PK target is somewhat foreign to
  current practice, at least outside the kinase
  area.
• Biochemical and cellular kinase assays need to
  be run in close conjunction with each other, even
  more so than for monovalent kinase inhibitors
  HCS technologies are beginning to impact
  optimisation in this way. Cellular assays can
  also measure inhibitory mechanisms which are
  missed by current biochemical methods
• Cross-target SARs are, by their nature, more
  complex than single-target SARs and compromises
  are generally to be expected
• Therefore it is very important to qualify these
  SAR compromises, preferably in cellular disease
  models or even primary cells

24
In Oncology MKIs will become the norm in the
kinase inhibitor field combination therapy and
MTDs with kinase and synergising non-kinase drugs
will emerge In Inflammation Certain MKIs will
make it to clinic and safety assessments will be
very interesting. For example, Palau have DD-2, a
dual Jak3/Syk inhibitor in preclinical for
autoimmune diseases and there are unpublished
data for related approaches Whither other
complex multifaceted diseases?
25

• Additional references
• Network pharmacology the next paradigm in drug
  discovery, A L Hopkins, Nature Chemical Biology,
  2008, 682.
• What does Systems Biology mean for drug
  discovery A Schrattenholz, Vukic Soskic, Current
  Medicinal Chemistry, 2008, 1520.
• Designed Multiple Ligands. An emerging drug
  discovery paradigm Richard Morphy, Zoran
  Rankovic, J Med. Chem., 2005, 6523.
• The physicochemical challenges of designing
  multiple ligands Richard Morphy, Zoran Rankovic,
  J Med. Chem., 2006, 4961.
• Logic models of pathway biology, Steven
  Watterson, Stephen Marshall, Peter Ghazal, Drug
  Discovery Today, 2008, 447.
• Can we rationally design promiscuous drugs A L
  Hopkins, J S Mason, J Overington, Current Opinion
  in Structural Biology, 2006, 127.
• Discovery of multitarget inhibitors by combining
  molecular docking with common pharmacophore
  features D Wei, X Jiang, L Zhou, J Chen, Z Chen,
  C He, K Yang, Y Liu, J Pei, L Lai, J Med. Chem.,
  2008, 7882.
• Selectively Nonselective Kinase Inhibition
  Striking the Right Balance R Morphy, J Med
  Chem.,.2010, 1413.

26
Acknowledgements for helpful discussions Rich
ard Morphy (Schering-Plough) Kate Hilyard,
Chris Newton (BioFocus) Ian James (Almac
Biosciences) John Overington (EMBL
Cambridge) Colin Telfer, Finbarr Murphy (Lee
Oncology)