Post on 03-May-2018
Jacob de Vlieg
Computational Drug DiscoveryComputational Drug Discovery
CDD subgroup, CMBI, November 15, 2006
Center of Molecular and Biomolecular Informatics (CMBI),Radboud University Nijmegen
Department for Molecular Design & Informatics (MDI)NV Organon
Key Goal CDD
To develop and apply state-of-the-art scientific in silico methods fordrug discovery
Two key areas:
• Structural (bio)molecular informatics (structure-based drug design)• BioInformatics for translational sciences
““ComputationalComputational Drug Drug DiscoveryDiscovery GroupGroup””
AgendaAgendaJacob de Vlieg:CDD introduction, some projects and scientific challenges in pharma
Sander NabuursA flexible approach to induced fit docking
Gijs SchaftenaarOverview CDD projects in Nijmegen
AssignmentWritten test (based on recent PhD study in CDD)
Ambition: top center for –problem driven- in silico drug discovery
Radboud University Radboud University
Academic Academic New scientific approachesNew scientific approachesTraining&educatioTraining&education
MDI Organon
ApplicationsExciting real life problemsMultidisciplinary NetworksInternal “wet” validation
Links to several academic groups & high tech companies
CDD: uniqueCDD: unique RadboudRadboud UniversityUniversity--OrganonOrganoncollaboration on collaboration on bioinformatisbioinformatis for drug discoveryfor drug discovery
The challenges at The challenges at PharmaPharma
• R&D expenditure is growing by 20% each year
• Average total cost per approved drug is ca. $880 Million and takes 12 years (Boston Consulting Group)
• Worldwide pharmaceuticals market: $240 billion in 2002 $3000 billion by 2020
Productivity and Innovation of PharmaProductivity and Innovation of Pharma
• Number of New Molecular Entities (NCE) is declining by 30% each year-Downward Trend (53 NMEs in 1996 … 21 in 2003 approved by FDA)-Most NMEs are ‘me too’ drugs; modulate targets for which drugs are on the market
-Only 2-3 NMEs address new protein targets
Drug discovery & development processDrug discovery & development process
Compounds optimized(wrt efficacy and safety) In model systems.
MarketResearch Developmentpreclinical
clinical
Failure rate over 80-90%
Does the compound work in man?
• Model systems in Discovery Research are insufficiently predictive for efficacy and safety in man ->Translational Sciences
• Difficult to find good starting molecules (drug-like leads) -> SBDD
very high and hardly changing attrition rates very high and hardly changing attrition rates
•• Data explosion of very complex and heterogeneous dataData explosion of very complex and heterogeneous data
•• Bioinformatics; scientific discipline at interface of biology, Bioinformatics; scientific discipline at interface of biology, chemistry, mathematics & ICTchemistry, mathematics & ICT
-- To put data into contextTo put data into context-- A new way of working based on sharing knowledgeA new way of working based on sharing knowledge
•• GenomicsGenomics in pharmain pharma-- Understand diseases at the molecular levelUnderstand diseases at the molecular level-- Hope for treatment of unmet medical needsHope for treatment of unmet medical needs
High expectations of genomics and High expectations of genomics and bioinformatics in pharmabioinformatics in pharma
….In reality number of new validated targets much less than expected
…Target validation very complex (i.e. to prove target is linked to disease
…Did genomics (and bioinformatics) fail?
Output of the linear, genomicsOutput of the linear, genomics--based based discovery process disappointingdiscovery process disappointing
Target discovery, validation & selection
Lead Optimization Output?Development
as “killing” fieldLead Discovery
Target validation & selection
TargetIdenti-fication
Lead Optimization
Pre & Early Clinical Development
Genomics-based profiling of compounds
Genomics-based profiling of compounds
Pharmacogenomics:Profile compounds in model systems
Pharmacogenomics:Profile compounds in model systems
Biomarkers for efficacyBiomarkers for efficacy
Toxicogenomics to predict adverse events
Toxicogenomics to predict adverse events
Pharmacogenetics for focused clinical trials
Pharmacogenetics for focused clinical trials
Did genomics fail? …Actually more genomics-based tools used across
entire drug discovery pipeline
Systems BiologySystems Biology
Lead Discovery
Protein familybased approach
Protein familybased approach
Integrated R&D to capitalize on genomics and informatics: the “hyperlinked’ organization
Exchange data and technologies* between discovery research and (exploratory) development
* IT systems, Bioinformatics&Genomics, ADMET, Safety Biomarkers, Imaging, Microdosing, Systems biology, PK/PD modeling, compound profiling, and so on.
Target validation & selection
TargetIdentification
Lead Optimization
Pre & Early Clinical Development
Lead Discovery
New technologies required to bridge R and DNew technologies required to bridge R and Dand to support PoC based drug discoveryand to support PoC based drug discovery
E.g. in silico technologies to support:
• knowledge sharing between Research and Exploratory Development
• Information flow between chemistry and biology • To fully capitalize on genomics, e.g.
– Integrated structure-based drug design– Systems biology– toxicogenomics for safety evaluation- Profiling methods (peptide recruitment)– biomarkers for efficacy– pharmacogenetics
• New technologies e.g. for non-invasive testing (Molecular Imaging, Microdosing,..)
DNA chip data analysisDNA chip data analysisBioinformaticsBioinformatics
ee--ADMETADMETTurning data in Turning data in
knowledgeknowledge
Virtual Screening & Virtual Screening & Library DesignLibrary Design
ClogP
Frequency
0
100
200
300
400
500
-8 -6 -4 -2 0 2 4 6 8 10 12 14
StructureStructure--based based drug designdrug design
Molecular Molecular DatabasesDatabases
In silico tools & databases usedIn silico tools & databases usedaccrosaccros entire entire genegene--toto--compoundcompound pipelinepipeline
Molecular Design & Informatics
Target discovery
Lead Discovery
Lead Optimization
Scientific Technology Focus at CDDScientific Technology Focus at CDD• In silico methods to produce (more drug like) leads
Biorange project de Vlieg, Schaftenaar, Folkertsma, Nabuurs et al.:Exploiting Structural Genomics Information To Incorporate Protein Flexibility In Drug DesignKeywords: Structure-based drug design, virtual screening, molecular information systems bridging chemistry and biology
• Bioinformatics to develop more predictive animal models by using comparative genomics and pathway data analysis
Biorange project Groenen, Hulsen, Fleuren, et al. :Protein Knowledge Building through Comparative Genomics and Data IntegrationKeywords: Pharmacogenetics, Translational Sciences
Several other Several other ““in silicoin silico”” projects at MDI/CDD projects at MDI/CDD interfaceinterface
• Integrative bioinformatics infrastructure (Schaik et al.)• Pharmacogenomics (Alkema et al) • Systems biology (Rullman)• Pharmacogenetics (Groenen et al.) • Toxicogenomics (Jan Polman )• Cheminformatics incl e-ADMET (Wagner and Ridder) • SBDD internal (Oubrie et al.) • Nuclear receptors (Lusher et al) • Kinase (Azevedo)• GPCR (Klomp) • Microarray platform (Bauerschmidt&Meijer)
(Connections to CDD; separate visit organized to Organon)
Goal: In silico methods to produce (more drug like) leads
Three key componentsA. Molecular Database Systems For Automatic Collection And
Interpretation Of Chemical And Biological Information; Binding site database (Folkertsma, Lusher et al.))
B. Computational Methods For Large Scale Analysis Of Binding-Site Architectures (Wagener et al.)
C. Techniques To Simulate Protein Mobility And To Predict Ligand-Induced Conformational States (Schaftenaar, Nabuurs et al.)
BioRange project 1: Exploiting Structural Genomics Information To Incorporate Protein Flexibility In Drug Design
(DeVlieg et al.)
Recent PhD study: Recent PhD study: Simon Folkertsma: Simon Folkertsma:
The Nuclear Receptor Ligand-Binding Domain; from biological function to drug design
A protein family based approach
A protein family based approach for nuclear receptors
Succesfully defended thesis at November 3
BioRange project 2:Protein Knowledge Building through Comparative
Genomics and Data Integration (Groenen et al.)
Key project componentsA. To establish an up-to-date repository of high quality (local) alignments of
protein sequences
B. Development of novel protein annotation methods
C. Knowledge integration of functional and structural data combined with data curation of alignments, primary annotation, classifications and visualizations.
Key goal: Bioinformatics to develop more predictive animal models by using comparative genomics and pathway data analysis
Wilco Fleuren: The evolution of the immunesytem from Chicken to Man; evaluating non-primate model sytems for auto-immune
diseases in drug discovery
Tim HulsenTesting statistical significance scores of sequence
comparison methods with structure similarityTim Hulsen1*, Jacob de Vlieg1,2 , Jack A.M. Leunissen3 and Peter Groenen21CMBI / NCMLS, Radboud University Nijmegen, The Netherlands2NV Organon, Oss, The Netherlands3Wageningen University and Research Centre Wageningen, The Netherlands
PhyloPat: phylogenetic pattern analysis of genesTim Hulsen1*, Jacob de Vlieg1,2 and Peter Groenen21CMBI / NCMLS, Radboud University Nijmegen, The Netherlands2NV Organon, Oss, The Netherlands
Tim finalises his thesis beginning of 2007
The ideal drug must have a specific biological action with no side-effects and no toxicity
Challenge in drug design is to optimize a molecule simultaneously on several -sometimes opposite- chemical and biological properties to be:
• effective• safe to use• chemically and metabolically stable,• synthetic feasible• bioavailable and high oral absorption• sufficiently unique to be patented
For example the right ‘solubility’ is dependant on a delicate balance:• the molecule has to be polar to cross the gastro-intestinal tract• to be sufficient lipophylic to penetrate cell membranes
BioRange 1:Key goal: computational methods to identify more
drug-like lead compounds
The long search for renin inhibitors by pharma: potent ligands (leads) were found but not drugs
Renin involved in regulation of blood pressure
Large binding region
Novartis
The The ReninRenin Story:Story:
Renin lead molecules too big lead optimization
LO process
•High binding affinity but insufficient…
•bio available •Low oral absorption•metabolic instable•and so on…
•Leads not “drug like”
Success dependant on drug likeness of lead compound!
More drug-like properties, but insufficient specific or affinity for renin
The role of protein flexibility in the The role of protein flexibility in the ReninRenin casecase
• Small ligand induces major conformational changes in the renin binding site
• Conformational changes not seen before in other ligand-renin X-ray structures
A very small A very small reninrenin--inhibitor discoveredinhibitor discovered
Small ligand: can be optimized into a drug
(lead is drug like)
Natural ligand (and analogs) too big to be drug-like
New lead creates its own specific binding site
(stabilizing a high-energy protein conformation)
previous leads
The role of protein flexibility in the The role of protein flexibility in the ReninRenin casecase
(a) (b) (c) (d)
(e) (f)
Recent paradigm describing proteins in a pre-existing ensemble of conformational states
(Ma et al.)
(a) (b) (c) (d)
(e) (f)
Recent paradigm describing proteins in a pre-existing ensemble of conformational states
(Ma et al.)
A moderate binder with preference for the conformation with the lowest free
energy and thus highly populated
(a) (b) (c) (d)
(e) (f)
Recent paradigm describing proteins in a pre-existing ensemble of conformational states
(Ma et al.)High-affinity binder that is specific for a less populated
conformational state.
• E.g. Set up MD simulation protocols to induce local, subtle unfolding of the protein at the surface/binding site
• Introduce “hydropobic” probes to find alternative binding site conformations or other binding (allosteric) sites
Develop in silico tools to find alternative binding sites and other relevant protein conformational states
Specific binding of drug to proteintarget involved in correct biological function
Binding to Anti-Targets (i.emembers of the same gene family, but with a different biological role) may result in adverseside effect
Grand challenge: drugs without adverse side effects: Grand challenge: drugs without adverse side effects: distinguish between target and antidistinguish between target and anti--targetstargets
target
drug
Anti target 1
Anti target 2
Anti target 3
KinaseKinase Gene FamilyGene Family
Kinases involved in crucial biological processes
However, up to app. 1995: “… kinases are non-druggabletargets …”
Selectivity problem: due to high sequence similarity and identical 3D structures of family members
Typically no or limited number of hits in existing HTS compound collections -> design of specific compound libraries for HTS based on structural information
Discovery of all kinases in human genome -> human kinasegene family encode (probably) 518 enzymes
Mapping of human kinome
Structural knowledge of many kinase-ligandinteractions at atomic detail
Active site of TARGET
Similarities and differences of kinase target and anti-target used to design specific drugs for autoimmune diseases:protein flexibility knowledge essential
Two structurally similar kinases involved in different Two structurally similar kinases involved in different biological pathways and responsesbiological pathways and responses
ANTI-TARGET
Biorange 1: component ABiorange 1: component ADrugs Drugs targetingtargeting NuclearNuclear ReceptorsReceptors
Pharmacologicaleffect
Collaboration on bioinformatics for drug discovery:
final biological or pharmacological effect
A DrugA Drug--Target interaction typically result into a cascade Target interaction typically result into a cascade of biochemical effects; a nonof biochemical effects; a non--linear processlinear process
Nuclear Receptor (NR)Drug A Nuclear Receptor (NR)Drug B
Drug (Drug (ligandligand) causes specific changes ) causes specific changes at at surface of thesurface of the Nuclear Receptor targetNuclear Receptor target
Binding of different CoActivator proteins result in different pharmacological effects
Subtle changes in the ligand/drug may cause very different pharmacological and side effects (non-linear process)
CoActivators
A different NR A different NR protein surfaceprotein surface results in results in recruitment of different corecruitment of different co--activators in the cellactivators in the cell
1. How can different interactions at the atomic level between compound and NR protein result in different, specific changes at the surface?
2. Can we predict the conformational changes and which co-activators will be recruited?
3. Can we find a link between compound structure, coactivator recruitment and pharmacological effect
Collaboration on bioinformatics for drug discovery:
Scientific ChallengesScientific Challenges
MCSIS – Molecular Class Specific Information System • System originally developed by prof. Vriend to understand protein mutation data*• To deal with complex structural data
* Horn et al; Nucleic Acids Res. 2001 29: 346-349 http://www.receptors.org
Radboud CMBI – Organon MDI collaboration:• Redesign the academic system to solve real life drug discovery problems• Automate the process & make it robust• Validate in silico methods by “wet” experiments
• Can we predict cross reactivity by finding essential interaction sites? • Can we reprofile drugs or reduce side effects of potential drugs
ApproachApproach
NURR1NURR1Nature (2003), 423, 555Nature (2003), 423, 555--560560
PXR SR12813
ERα THC
NR0B1 DAX1 VALL.VNR0B2 SHP VALL.VNR1A1 THA FIMALHNR1A2 THB1 FIMALHNR1B1 RRA FSLIFGNR1B2 RRB FALIFGNR1B3 RRG1 FALIFGNR1C1 PPAR CSILVHNR1C2 PPAS CTILVHNR1C3 PPAT CSILIHNR1D1 NRD1 FFLFFHNR1D2 NRD2 FFLFFHNR1F1 RORA CIVMFHNR1F2 RORB CIVMFHNR1F3 RORG CLVMFHNR1H2 NRH2 FAMTFHNR1H3 NRH3 FAMTFHNR1H4 NRH4 LAMSGHNR1I1 VDR LVISCHNR1I2 PXR MSCFCHNR1I3 NRI3 FNCLCYNR2A1 HN4A VMLALMNR2A2 HN4G VMLALMNR2B1 RXRA IALFLCNR2B2 RXRB IALFLCNR2B3 RXRG IALFLCNR2C1 TR2 IAFLAINR2C2 TR4 IAFLSINR2E1 NR21 VAFILTNR2E3 NR23 ISFALRNR2F1 COT1 IAFAAVNR2F2 COT2 IAFAAVNR2F6 EAR2 VAFAALNR3A1 ESR1 LALLFGNR3A2 ESR2 LALLFGNR3B1 ERR1 LFLVFVNR3B2 ERR2 LALIYANR3B3 ERR3 LALVYANR3C1 GCR LGMLFLNR3C2 MCR LASLFLNR3C3 PRGR LGMLFLNR3C4 ANDR LGMMFLNR4A1 NR41 FLFLFGNR4A2 NR42 FLFLFGNR4A3 NR43 FLFLFGNR5A1 STF1 LALHLANR5A2 NR52 MALHFANR6A1 NR61 LAISAM
48 NRs in 48 NRs in human human genomegenome
discovereddiscoveredNRNR--ligandligand
XX--ray Complexesray Complexes
How do we connect this structural profile to How do we connect this structural profile to relevant changes at the relevant changes at the protein surfaceprotein surface??
• X-ray surface data of complexes
• Organon: measure binding profiles of > 80 recruitedcoactivators (peptides)
– Indirect “Image” of the protein surface– Biological relevant information (wet validation)
Recruited peptideX1
X2
X3…X80
9-cis RA
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
NC
oR_1
NC
oR_3
NC
oR_4
SMR
T_ID
1
PR_H
12
BN2
HR
_1
LCoR
HR
CoA
_2
SRC
1a_4
SRC
2_1
SRC
2_3
SRC
3_2
CBP
_1
RIP
140_
3
RIP
140_
6
RIP
140_
8
PGC
_1
TRAP
220_
1
ARA7
0
C33
Ppt4
-1
SHP_
1
DAX
_3
ARAF
1
LXR
a_H
12
9-cis RA
Peptide no.
Affinity
Binding affinity
Peptide recruitment binding studies Peptide recruitment binding studies Biological relevant image of surface changeBiological relevant image of surface change
9-cis RA
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
NC
oR_1
NC
oR_3
NC
oR_4
SMR
T_ID
1
PR_H
12
BN2
HR
_1
LCoR
HR
CoA
_2
SRC
1a_4
SRC
2_1
SRC
2_3
SRC
3_2
CBP
_1
RIP
140_
3
RIP
140_
6
RIP
140_
8
PGC
_1
TRAP
220_
1
ARA7
0
C33
Ppt4
-1
SHP_
1
DAX
_3
ARAF
1
LXR
a_H
12
9-cis RA
oleic acid
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
NC
oR_1
NC
oR_3
NC
oR_4
SMR
T_ID
1
PR_H
12
BN2
HR
_1
LCoR
HR
CoA
_2
SRC
1a_4
SRC
2_1
SRC
2_3
SRC
3_2
CBP
_1
RIP
140_
3
RIP
140_
6
RIP
140_
8
PGC
_1
TRAP
220_
1
ARA7
0
C33
Ppt4
-1
SHP_
1
DAX
_3
ARAF
1
LXR
a_H
12
oleic acid
+ +
+
Co-factor peptidesNR-protein
+Compound Peptide binding profile
O
OH
HOO
Spotfire visualization to link complex structural profiles, peptide recruitment and
pharmacological profiles
Profile Chart
0
5
10
15
20
H3 H4 H5 H7 H10 H12
NR - 6HIS~Era~fdom
-1.5
-1
-0.5
0
0.5
1
Peptide recruitmentprofile
Cellular data
Animal data
Correlate with animal data
Correlate with cellularTrans-activation data
Structural ligand-protein profile
PCA
PCA