COMPUTATIONAL DRUG DISCOVERY: INTEGRATING A COLLABORATIVE DATABASE Sean Ekins MSc, PhD, DSc Collaborations in Chemistry, Jenkintown, PA. Department of Pharmacology, University of Medicine & Dentistry of New Jersey-Robert Wood Johnson Medical School, Piscataway, NJ. School of Pharmacy, Department of Pharmaceutical Sciences, University of Maryland, Baltimore, MD.
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
COMPUTATIONAL DRUG DISCOVERY: INTEGRATING A COLLABORATIVE DATABASE
Sean Ekins MSc, PhD, DSc
Collaborations in Chemistry, Jenkintown, PA.Department of Pharmacology, University of Medicine & Dentistry of New Jersey-Robert Wood
Johnson Medical School, Piscataway, NJ.School of Pharmacy, Department of Pharmaceutical Sciences, University of Maryland,
Baltimore, MD.
Examples of informatics technologies
Infrastructure and other underlying toolkits: ChemAxon, Daylight, Spotfire, OpenEye, Pipeline Pilot, MOE, Oracle, MySQL
Modeling Software: CTC labs, Accelrys, Tripos, Partek, Hypercube
Registrations systems: Accord, ChemOffice, IDBS, ISIS, Collaborative Drug Discovery (CDD),
Chemistry Databases: Beilstein, SPRESI, LUCIA, ACD, ACX, Scifinder, Derwent, Gene-family SAR Databases (GVK, Jubilant-Biosys, Eidogen-Sertanty), Prous’ Drug Data Report from Symyx, Pergamon’s Comprehensive Medicinal Chemistry etc..
Hohman et al, submitted 2008
Examples of technologies leveraging or modeling network effects
Knowledge, Information and Social Networking: Wikipedia, OpenWetWare, BioSpace, BioPortfolio, ACS Member Network, LabMeeting, Laboratree, SciLink, SciMeet, Nature Network, Ensembl
Service and Product Provider Databases: Assay depot, R&D Chemicals
Drug Discovery Platforms: Collaborative Drug Discovery (CDD), SEURAT, (Synaptic Science LLC), NextBio
Systems Biology Tools: Ariadne Pathway Studio, Cytoscape, Ingenuity Pathways Analysis, MetaCore, MetaDrug, WikiPathways, Systems Biology Research Tool, (also see http://www.biochemweb.org/systems.shtml)
Federated Databases: ChemSpider, eMolecules, ZINC
Public Content Databases: KEGG, NCI, PDSP, PubChem, ChemBank, (see also http://depth-first.com/articles/2007/01/24/thirty-two-free-chemistry-databases), SureChem
Hohman et al, submitted 2008
• PXR increases transcription of CYP3A4 and >36 other genes
Drug metabolismDrug transport
The Pregnane-X-Receptor
PXR
PXR agonistsBile saltsCholesterol metabolites
StatinsPPAR antagonistsCalcium channel modulatorsSynthetic peptide mimeticsAnticancer compoundsHIV protease inhibitors
Herbal componentsCarotenoidsVitamins
Endocrine disruptersPesticidesPlasticizers
and many more……………..
Endogenous
Drugs
Exogenous
Environmental Contaminants
PXR Agonist Machine Learning and Docking Comparison
We have compared different methods for predicting agonist binding to human PXR:
Training set 98 hPXR activators and 79 hPXR non-activators (Ung et al., Mol Pharmacol 71:158-168 2007) Recursive partitioning (RP) Random forest (RF) Support Vector Machines (SVM) Using VolSurf 3D descriptorsDocking (FlexX)
Large external test set N = 145 molecules (82 active, 63 inactive)
Khandelwal et al., Chem Res Toxicol, 21(7):1457-67 (2008)
SVM = 66.9% correct classificationComparable test molecule space vs training set with PCALatest Work compares GOLD docking and custom scoring functionsPredictions uploaded in CDD
Pioglitazone docked in 1NRL
Machine Learning is Better than DockingDocked into 4 different ligand co-crystallized structures of hPXRwith hyperforin (PDB ID: 1M13), rifampicin (PDB ID: 1SKX), T0901317 (PDB ID: 2O9I) and SR12813 (PDB ID: 1NRL).
Khandelwal et al., Chem Res Toxicol, 21(7):1457-67 (2008)
PXR Ligand Database using Collaborative Drug Discovery
PXR antagonist drug discovery
Cancer drugs act as PXR agonists, increasing own metabolism and transport out of cellsHow could we block this?Preferably find a clinically used drug?
PXR Antagonist PharmacophoreCompounds can “switch off” PXR 3 azoles shown to antagonize PXR ~ equipotent (10-20μM) mutagenesis data indicates they bind outer surface of PXR – AF-2 binding pocket
Can a pharmacophore infer features needed to antagonize hPXR?
Ekins et al., Mol Pharmacol 72:592–603, (2007)
Huang et al., Oncogene 26: 258-268 (2007), Wang et al., Clin Cancer Res 13: 2488-2495
Hydrophobe / ring aromatic
H-bond acceptors
Antagonists require a balance between hydrophobic and hydrogen bonding features.
PXR Antagonist Binding Site/s - Docking
Ekins et al., Mol Pharmacol 72:592–603, (2007)
2 separate binding sites on either side of Lys277- identified with GOLD rigid docking in 1NRL chain A
azoles would interfere with SRC-1 binding in the AF-2 site. One site is predominantly hydrophobic -15 amino acids.
Lys277 most likely serves as a “charge clamp” for interaction between the co-activator SRC1 (His687) and PXR
Azoles compete with SRC-1 for AF-2
Piperazine etc may not be necessary- Solvent exposed
Screened four databases – known drugs and commercially available molecules, N = 3533 67 hits retrieved We tested in vitro a small number based on their pharmacophore fit values and mapping to the pharmacophore features Followed up hits with similarity searching using ChemSpider.com, emolecules.com
PXR Antagonist Database Searching
Ekins et al., Mol Pharmacol, 74(3):662-72 , (2008)
SPB00574 2.14 24.8
SPB03255 2.22 6.3
Catalyst fit IC50 (μM)
PXR Antagonist Database Searching Finds New Hits
Further similarity searching retrieved 4 active analogs of SPB03255 Also tested leflunomide
IC50 = 6.8 μM, Ligand efficiency >100% higher than ketoconazole
Ekins et al., Mol Pharmacol, 74(3):662-72 , (2008)
Ketoconazole
SPB03255
PXR GOLD Docking in AF-2 Domain – Smaller May Be Better
Side View Front view
Ekins et al., Mol Pharmacol, 74(3):662-72 , (2008)
How could the PXR antagonist work benefit from CDD?Currently multiple labs (CIC, UMDNJ, BMS, Albert Einstein) collaborating on project only CIC uses CDD…We could have used CDD to show that leflunomide is not a promiscuous inhibitor, and has been tested vs malaria and TBAlso indicates some Asinex compounds of moderate similarity may be worth testingCould upload ligand efficiency data
Accelerating Computational Drug Discovery Using A New Collaborative Database and QSAR Models
Kip Guy, St. JudeMalaria biology, screening
Alex Tropsha Lab, UNCComputational Chemistry, informatics
Malaria data, global
Liying Zhang Hao Zhu
Goal: Accelerate computational drug discovery for malaria
Bunin, Ekins etc Facilitators
Collaboration in 2008 over a few months
Malaria Case Study
Uploaded in vivo data on over 12,000 molecules from the 61-year-old, two-volume “A Survey of Malaria Drugs,” edited by Frederick Y. Wiselogle(1946). This database integrates 35 protocols for in vivo antimalarial activity screening, different in types, labs, etc
Chemistry space of in vitro anti- P. falciparum activity data
Workflow for k-nearest neighbors classification models
QSAR Models have high prediction accuracy
70 compounds in the 429 compounds were randomly selected as external validation set.QSAR kNN (k-nearest neighbor) modelsResults:
More than 300 acceptable QSAR classification models.True negative:83.7% True positive: 72.4%Total prediction accuracy: 81.4%
The most frequently used fragment descriptors for the in vitro models
Antimalarial hit identification
based on virtual screening of commercial chemical databases with the in vitro kNN model
A Tropsha Group in Silico Protocol Example in CDD
My Perspectives from a small virtual pharma
Considerable amounts of outsourcingIn vitro data and in vivo data may come from different groups – frequently on different continentsCompany may have few staff working remotelySharing of excel, word and pdf filesKeeping track of compound information grows more difficult with timeNeed a single place that everyone can assess from wherever they are locatedNeed a way to partner or find partners from bigger companies
My Perspectives from academic-pharmacollaborations (computational- in vitro)
Loose collaborationSignificant gaps of time between suggesting molecules, purchase, testing and return of dataNeed for a bridge at least between the academic groups to share molecule ideas and input of in vitro dataNeed for a means to find potential additional collaborators to expand data on molecules
Why collaboration is important
Scientific collaboration networks
OpenWetWare, BioSpace, BioPortfolio, ACS Member Network, LabMeeting, Laboratree, SciLink, SciMeet, Nature Network, EnsemblCommunity of Science, LinkedIn, Google, email, telephone!!
What is missing ? - data, scienceWhy don’t all these networking sites connect ?
Flavors of collaborationFocused
Select a collaborator to do a specific taske.g. you are an experimental scientist but need a computational chemist for a projectI have been the comp-chemist in such collaborationsThese may be short term for one paper or grant
PromiscuousYou have a wide array of projects and need others to assist with them
You look for independent but willing partnersYou may have many of these collaborations over the yearsGenerally your publications will be multi-author and they will have little overlap in topic or contentI have been a collaborator on such projects
BroadSelect a collaborator as a trusted partner
e.g. you may share common interests and these progress beyond one project These may be multi year projects papers or grantsMay grow organically depending on interests of each person e.g. each collaborator brings in their own collaboratorsI am in a couple of such collaborations
Types of cross-organization collaboration
Academic to academiccompetition concerns e.g. grant fundinguseful for career networkinguseful to share costs as science becomes more expensive
Academic to industryIP concernspublication concerns from both sides$ concernsethical concerns
Industry to industrySmall pharma may be looking for a bigger partner for their molecule/sBigger pharma may be looking to outsource workBigger pharma may be looking to outlicense/inlicensecompoundsPatent concerns$ concerns
What makes a good collaborator/ collaboration
Decide who does what at the outset and stick to itShare, share, shareGive and take (Respect!)Maintain good lines of communicationAcknowledge
How can I find a good collaborator??
Vicens Q, Bourne PE (2007) Ten simple rules for a successful collaboration. PLoS Comput Biol 3(3):e44. doi:10.1371/journal.pcbi.0030044
How can we foster collaborations via CDD?
ManuallyIts not only what you know but who..Slow, hit or miss
AutomatedNeed to find compatabilityVia networking software or other algorithmsShare your interestsShare your needsProvide a profileProvide a feedback rating of your collaborators? Who do you want to work with
Matt KrasowskiUniv Pittsburg
Erin SchuetzSt Jude
Sridhar ManiAlbert Einstein
Konstantin BalakinChemDiv / Russia
Peter SwaanUniv Maryland
Joe PolliUniv Maryland
me
Mike SinzBMS
Bill WelshUMDNJ & Snowdon Inc
Sandhya KortagereDrexel
Renee ArnoldACT LCC Barry Bunin
CDD Inc
Ken BachmannUniv Toledo
Evan KharaschWashington Univ
Bing YanSt Jude
Alex TropshaUNC
Liying Zhang, Hao Zhu, Denis Fourches, (UNC), Kip Guy (St Jude), Erica J. Reschly, Manisha Iyer (University of Pittsburgh), Vladyslav Kholodovych, Ni Ai, Dima Chekmarev, (UMDNJ), Akash Khandelwal, (University of Maryland), Rachana Patel, (University of Toledo), Ingenuity, Accelrys, Moses, Kellan et al., (CDD), www.chemspider.com
Collaborator Groups:
This talk
Related Documents
