Embed Size (px)
Citation preview
Workflows supporting drug
discovery against malaria
Barry Hardy, Thomas Exner and Alessandro Contini
ACS, Boston
17 August 2015
Islands – a reality of geography
Source: Baily Ed, U.S. Fish and Wildlife
Service
Knowledge-oriented Framework
Based on Nonaka & Takeuchi, The Knowledge
Creating Company, 1995
Socialisation Externalisation
Internalisation Combination
Tacit
Tacit Tacit
Tacit
Explicit Explicit
Explicit
Explicit
Knowledge
Sharing:
Discussions
Knowledge Creation from R&D: Data, Codes
Learning:
Apply Models
Knowledege Combination: Predictive Models
Acceptance
Community of Practice starting point
We started with
community
development
interactions, both
virtual and face-to-
face...
Knowledge Cafe Discussions
Community of Practice starting point
...and continued by holding
workshops and InterAction Meetings
in US and Europe
Bryn Mawr College, US
Oxford University, UK
Workshop Activity
Workshop Activity
Hands-on activity – early Oxford workshop
Methods Software Examples Case Studies
Malaria Treatment: Lack of Investment
• 2.5 billion people at risk
• 500 M cases yearly
• ca. 1 M deaths yearly
• Many child fatalities
• Brain Damage, Impaired Development
• Few drugs, no vaccine yet
• Impact on Education, Community, Income,
Conservation, Sustainability
Tropical diseases: 13
Malaria: 3
Total: 1223
Greenwood & Mutabingwa, Nature 415:670-672
Drug development outcome, last quarter of the XXth century
Drug Resistance is a challenge
Emergence and spread of chloroquine resistance
New emergence in SE Asia of Artemisinin resistance
Drug Resistance is a challenge
New emergence
in SE Asia of
Artemisinin
resistance
Erythrocytic schizogony: ► Pfmap-2
► PfCDPK1Winzeler
► PfPK7 ► Pbcrk-1 ► Pfnek-1 ► PfPK5 ► PfPK6 ► PfPK9 ► PfCK2 ► PfGSK3 ► PfTKL1 ► PfTKL3 ► Pfcrk-3 ► Pfcrk-4 ► PfARK1 ► PfARK2 ► PfARK4 ► PfPK4
Non-essential for erythrocytic schizogony ►Pfmap-1 ►PfPK7
► PbCDPK3Ishino, Billker
► Pfnek-4 / Pbnek-4
► PbDCPK4Billker ► Pbmap-2
► PfPKGBaker
► Pfnek-2 ► Pfnek-3 ► Pfnek-4 ► Pfcrk-5 ► PfeIK1 ► PfeIK2 ► PfTKL-2 ► PfTKL-4 ► PfTKL-5
Gametogenesis: ►PbDCPK4Billker ►Pbmap-2
►PfPKGBaker
Ookinete migration: ►PbCDPK3IIshino, Billker
Oocyst maturation: ►PfPK7
Ookinete maturation: ►Pbnek-4 ►Pbnek-2
Sporozoite infectivity ►PbCDPK6coppi
Plasmodium Life Cycle & Kinome
Plasmodium Life Cycle & Kinome
Baldauf, Science, 2003
Plasmodium
Plasmodium Life Cycle & Kinome
STE
CAMK
AGC
TK TKL
CK1
CMGC
+OPK
CSNK1G1 PF11_0377 (PfCK1) CSNK1D PFB0150c MAP2K1 MAP3K3 MAP4K1 PAK1
PF14_0320 PF11_0220 MAL6P1.191 PF13_0258 TGFbR1 PFB0520w ZAK LIMK1 RAF1 SRC INSR EGFR PFI1280c PFI1290w MAL6P1.146 (PfPK4) PfA0380w PF14_0423 PFL0080c MAL6P1.56 PFL1370w (Pfnek-1) NEK7
PFE1290w NEK1 MAL7P1.100 (Pfnek-2) MAL7P1.91 (PfEST) PF11_0488 MAL13P1.196 PFC0755c (Pfcrk-4) MAL6P1.271 (Pfcrk-5) PF13_0206 (PfPK6) PFD0740w (Pfcrk-3) MAL13P1.279 (PfPK5) CDK2 PF10_0141 (Pfmrk) PFD0865c (Pfcrk-1) PF11_0147 (Pfmap-2)
PF14_0294 (Pfmap-1) MAPK1 PF11_0096 (PfCK2)
PF08_0044 (PfPK1) MAL13P1.84 PFC0525c (PfGSK3) GSK3B PF11_0156 PF14_0431 (Pflammer) CLK1 PFC0105w PF14_0408
PF14_0392 MAL7P1.73 PFL1885c (PfPK2) PF11_0242 PFB0815w (PfCDPK1) PF07_0072 (PfCDPK4) PF13_0211 MAL6P1.108 (PfCDPK2) PFC0420w (PfCDPK3) MAPKAPK2
CAMK2A CAMK1 DAPK1 PF11_0239 PF14_0227 PFC0485w PFB0665w MAL7P1.18
PF11_0060 PF11_0464 PFC0385c MAL13P1.278 PF11_0227 PF14_0346 (PfPKG) ROCK1 PFL2250c (PfPKB) PRKCA AKT1 PRKACA PFI1685w (PfPKA) PF14_0516 (PfKIN) PF14_0476 PF13_0085 PfI1415w PFB0605w (PfPK7) PFL2280w
PlasmoDB mining
(HMM, BLAST, Gene Ontology prediction)
Alignment with selected human PKs, manual edition
Phylogenetic tree
• 518 protein kinases (2% of genes) • Involved in essentially all cellular processes (30% of the proteome is phosphorylated)
Network
Virtual Organisation
Opportunity
Call for Tender
Need for joint effort
Major project
Coordinat
or
Partner 1
Partner 2
Partner 3
Partner 4
Partner 5
Partner 6
Partner 7
Applied to Scientists Against Malaria (SAM)
Creation of VO from Collaboration Pool
Scientists Against Malaria Founding Partners
• Barry Hardy & Roman Affentranger
(Douglas Connect)
• Alessandro Contini
(University of Milan)
• Hugo Gutierrez de Teran
(Public Galician Foundation of
Genomic Medicine)
• Jeffrey Wiseman & Matt Clark
(Pharmatrope)
• Jeff Spitzner (Rescentris)
• Ruben Papoian, William Seibel &
Sandra Nelson (Univ. of Cincinnati
Drug Discovery Center)
• Sharon Bryant (Inte:Ligand)
• Andrew Wilks & Isabelle Lucet
(Monash University)
• Christian Doerig Coordinator of the
FP7 MALSIG project on signalling in
Malarial parasites
• Matteo Dal Peraro (EPFL, Lausanne)
www.scientistsagainstmalaria.net
Hardy, B &Affentranger, R, Drug Discov Today. 2013 Jul;18(13-14):681-6
Workflow Infrastructure for SAM
Screened
Library
Refine
Predictions
CERF
Data
Data
Toxicity Predictions
Toxicity
Assays
Decision
Dashboard
(Safer)
Drug Leads
Hardy, B &Affentranger, R, Drug Discov Today. 2013 Jul;18(13-14):681-6
Target Model
• Starting point – no protein structure, no known ligand/inhibitor
• Initial model of PfMAP2 Kinase Protein built based on existing knowledge
• Library of Potential Inhibitors created (ca. 1.2 M structures)
• Virtual screening runs carried out at computing centres in Italy, Spain and USA
• Protein expressed in Monash, Australia and shipped to screening centre at Univ. Cincinnati where assays were developed and run
Binding Pocket
Pharmacophore-based Screening
Dock Screening Prediction
Resolving Inconclusives
Mo
de
l 1
Mo
de
l 2
Mo
de
l 3
1 0 1
Mo
de
l 1
Mo
de
l 2
Mo
de
l 3
1 0 1
Ass
ay 1
Ass
ay 2
Ass
ay 3
- - -
Ass
ay 1
Ass
ay 2
Ass
ay 3
- - - Recommendation Rules:
0 0 1
0 1 1
1 0 0
1 0 1
1 1 0
0 1 0
Hit, high confidence
Not a hit, high confidence
Inconclusive results, further study needed
ELN
Synergy
OpenTox
1 1 1
0 0 0
Resolving Inconclusives
Mod
el 1
Mod
el 2
Mod
el 3
- - -
Mo
de
l 1
Mo
de
l 2
Mo
de
l 3
1 0 1
Mo
de
l 1
Mo
de
l 2
Mo
de
l 3
1 0 1
Ass
ay 1
Ass
ay 2
Ass
ay 3
- - -
Ass
ay 1
Ass
ay 2
Ass
ay 3
- - - PM
ELN
Synergy
OpenTox
!
?
Resolving Inconclusives
Mod
el 1
Mod
el 2
Mod
el 3
- - -
Mo
de
l 1
Mo
de
l 2
Mo
de
l 3
1 0 1
Mo
de
l 1
Mo
de
l 2
Mo
de
l 3
1 0 1
Ass
ay 1
Ass
ay 2
Ass
ay 3
- - -
Ass
ay 1
Ass
ay 2
Ass
ay 3
- - -
Ass
ay 1
Ass
ay 2
Ass
ay 3
- 1 -
Ass
ay 1
Ass
ay 2
Ass
ay 3
- 1 1
ELN
Synergy
OpenTox
Event-driven Collaboration Dashboard
www.scientistsagainstmalaria.net
Hardy, B &Affentranger, R, Drug Discov Today. 2013 Jul;18(13-14):681-6
Preliminary Results
Pharmacophore Search
Found 696 fits in library of >300,000 compounds
Evaluated energies by free energy simulations
Docking predictions
996 compounds predicted as consensus between three docking screens (AutoDock, Vina, Glide)
Binding Assay
Assay required Development & Optimisation Time
Several micromolar hits with dose-response curves obtained from initial screenings.
Example hit:
AC50=1.735µM
Incorporation of Holistic Predictive ADME & Toxicity
Models
Predictive toxicology model building
Ass
ay 1
Ass
ay 2
Ass
ay 3
- - -
Mo
de
l 1
Mo
de
l 2
Mo
de
l 3
- - -
Mo
de
l 1
Mo
de
l 2
Mo
de
l 3
- - -
Ass
ay 1
Ass
ay 2
Ass
ay 3
- - -
Mo
de
l 1
Mo
de
l 2
Mo
de
l 3
- - -
Ass
ay 1
Ass
ay 2
Ass
ay 3
- - - M
od
el 1
Mo
de
l 2
Mo
de
l 3
1 - - M
od
el 1
Mo
de
l 2
Mo
de
l 3
1 0 - M
od
el 1
Mo
de
l 2
Mo
de
l 3
1 0 1
Algo-rithms
Data
ELN
Synergy
OpenTox
Predictive toxicology model building
Ass
ay 1
Ass
ay 2
Ass
ay 3
- - -
Mo
de
l 1
Mo
de
l 2
Mo
de
l 3
- - -
Mo
de
l 1
Mo
de
l 2
Mo
de
l 3
- - -
Ass
ay 1
Ass
ay 2
Ass
ay 3
- - -
Mo
de
l 1
Mo
de
l 2
Mo
de
l 3
- - -
Ass
ay 1
Ass
ay 2
Ass
ay 3
- - - M
od
el 1
Mo
de
l 2
Mo
de
l 3
1 - - M
od
el 1
Mo
de
l 2
Mo
de
l 3
1 0 - M
od
el 1
Mo
de
l 2
Mo
de
l 3
1 0 1
ELN
Synergy
OpenTox
Mo
de
l 1
Mo
de
l 2
Mo
de
l 3
1 0 1
Outline of Current Screening Approach
Outline of Extended Screening Approach
Including Tox Prediction
Protein-Ligand ANT System
O. Korb, T. Stützle, T.E. Exner
• Dorigo (1992)
• Looking at real ant colonies
(Iridomyrmex humilis)
– Single ant: limited
– Colony: highly efficient
• One version of Swarm Intelligence
• Ants solve shortest-path problems when searching for food
• Indirect communication via pheromone:
– Ants deposit pheromones while working
– Ants follow pheromone trails
Ant Colony Optimization
• path protein-ligand conformation
• short path low energy protein-ligand conformation
• long path high energy protein-ligand conformation
SCORE: -106 SCORE: -68
ACO - Problem Analogies
Kinases binding site flexibility
Alessandro Contini – eCheminfo2015, Ahmedabad, January 04-09
Improving docking results - target structure?
Homology model (green) vs crystal structure (3NIE.pdb, blue)
the loop fragment Cys156-Ser166 (magenta) was removed due to poor
information in templates
Different homology models
m1: tmpl MapK1+82A,
Loop deleted
m1: tmpl MapK1+82A,
Loop not deleted
m1: tmpl MapK1+ATP,
Loop not deleted
SW: Plants
Test set: 5 actives, 60 inactive (20 high mw, 20 mid mw, 20 low mw); standard
protonation; tautomers
homology model vs crystal structure (3NIE.pdb, chain B)
SW: Plants
Test set: 5 actives, 60 inactive (20 high mw, 20 mid mw, 20 low mw); standard
protonation; tautomers
Docking Method?
PLANTS VINA
SW: Plants or Vina
Test set: 5 actives, 60 inactive (20 high mw, 20 mid mw, 20 low
mw); standard protonation; tautomers
Ligand Efficiency
Scoring functions like ChemPLP favor larger molecules due to the
larger amount of attractive interactions
Ligand Efficiency = ΔG/natoms= Score/natoms
Renormalization (MW, N°heavy atoms…) might help
PLANTS: crystal vs homology model
Crystal structure Homology model
SW: Plants
Test set: 5 actives, 60 inactive (20 high mw, 20 mid mw, 20
low mw); standard protonation; tautomers
PLANTS vs VINA (normalized*)
PLANTS VINA
*Vina SNH: 4*(VinaEnergy / (Heavy Atom count / 3))
PLANTS: Total Score vs Heavy Atom Normalization
SW: Plants
Test set: 5 actives, 60 inactive (20 high mw, 20 mid mw, 20 low mw);
standard protonation; tautomers
Data provided that is relevant for this project: • Activity against (growth inhibition of) P. falciparum strain 3D7 (common strain) • Activity against (growth inhibition of) P. falciparum strain DD2 (multi-drug resistant strain) • Cytotoxicity against (growth inhibition of) human hepatocytes, HepG2 (hepatoma cells)
“Malaria Box”*: A collection of chemical compounds active against (i.e. inhibiting
growth of) the malaria parasite Plasmodium falciparum
5267
8086
Antimalarial Activity
3D7 and DD2 3D7 only
10216
3137
Cytotoxicity
Not Toxic Toxic1339 active* but toxic
3928
active*
and
non-toxic
1789
inactive*
and
toxic
*Gamo et al., Nature 465(7296), 305-10 (2010)
*) against P. falciparum DD2
TCAMs Malaria Box – Predictive Toxicology Workflow Development & validation
Mining Human Adverse Events
Adverse Event Groups Group Name Hepatic function abnormal
FuAbn Liver disorder
Hepatic necrosis Nec Cytolytic hepatitis
Hepa Hepatitis Hepatitis acute Hepatitis toxic Cholestasis
CholJa Jaundice Hepatitis cholestatic jaundice cholestatic Yellow skin Hepatic failure
HepFail Hepatitis fulminant Acute hepatic failure Hepatorenal failure Hepatotoxicity
HepTox Hepatomegaly Hyperbilirubinaemia Hepatosplenomegaly
Combination Rule for Event Groups: Associate a drug with a group if the sum of individual event values is larger of equal to 0.4.
Event-drug pair values in Titanium Predictions: 0 : no association (0) 0.35-0.4 : non-significant association (0) > 0.4 : significant association (1) Combination Rule for Event Groups:
Associate a drug with a group if the sum of individual event values is non-zero
Event-drug pair values in Titanium Data: 0: no association 1: significant association
Adverse Events Model Development
Combination Rule for Event Group Predictions: Associate a drug with a group if either the Pharmatrope or the Leadscope predcition is positive (or both)
AERS Consensus: Count the number of Adverse Event Group Consensus associations. If more than one is positive, the AERS Consensus is positive.
OpenTox Consensus: Negative if both carcinogenicity and the micronucleus assay predictions are negative, OR if the Cramer Rule classification is Class I. Positive otherwise.
TCAMS Cytotoxicity: Positive if > 30% growth inhibition at 10 µM.
TCAMS Antimalarial Activity: Positive if > 80% growth inhibition of P. Falciparum DD2 at 2 µM.
Combining Predictions & Experimental Data
~5300 compounds active against DD2
Cytotoxic “Safe” (74.8%)
AERS Predictions
Events Associated
TCAMS Data
~3900 compounds active against DD2 and not cytotoxic
OpenTox Predictions
410 compounds active against DD2, not cytotoxic and no adverse events
predicted
“Safe” (10.4%)
Toxicities predicted
“Safe” (7.1%)
29 compounds without any indication of adverse events or
toxicities
Compound Prioritisation Filtering
TCMDC-131287: -No predicted association with adverse events (consistent)
-Negative for carcinogenicity and mutagenicity
-No inhibition of HepG2 growth -Strong inhibition of P. falciparum DD2 growth
TCMDC-138057: -Predicted association with many adverse events groups -Positive for carcinogenicity and mutagenicity -Considered safe (Class I) with Cramer rules -Inhibition of HepG2 growth could not be measured -Strong inhibition of P. falciparum DD2 growth
TCMDC-125641: -No adverse event association predicted with Pharmatrope models -Strong association with all five adverse events groups predicted with the Leadscope models
-Negative for carcinogenicity and mutagenicity - Intermediate inhibition of HepG2 growth (33% at 10 µM) -Strong inhibition of P. falciparum DD2 growth (100% at 2 µM)
“Safe” Ambiguous, Further Data Required
“Toxic”
TCMDC-137245: -Associated with 4 and 5 (out of 5) adverse events group by Pharmatrope and Leadscope, respectively
-Positive for carcinogenicity and mutagenicity, Cramer Class III
-67% HepG2 inhibition (10 µM) -91% P. falciparum DD2 growth inhibition (at 2 µM)
Example Classified Compounds
Workflows supporting drug
discovery against malaria –
thanks for your attention and …
… consider joining our efforts!
Barry Hardy, Thomas Exner and Alessandro Contini
Contact:
Barry dot Hardy –at/a douglasconnect dot com
ScientistsAgainstMalaria.net
