USUGM 2014 - Gerald Wyckoff (Chemalytics): Development of the Chemalytics Platform for Drug...
Transcript of USUGM 2014 - Gerald Wyckoff (Chemalytics): Development of the Chemalytics Platform for Drug...
What drives our research?
The pharmaceutical industry is facing spiraling drug development costs
while R&D productivity remains stalled
6 of the 10 highest-grossing branded products will or have lost patent
exclusivity this year (2014)
Reuters notes that the industry spent $65 billion on drug R&D in the U.S. in
2009, but approval rates have sunk 44% over the past 13 years
Background
Importance of identifying valid targets and therapeutic compounds
Tools currently in use:
Structure-based virtual screening
Receptor-based virtual screening
Other computational tools
Drawbacks to current implementation of high-throughput virtual screening:
Computationally intensive
Limited access due to high cost of infrastructure
Solution:
Virtual screening in the cloud
Provides computational resources scalably and only when needed
Chemalytics Platform
Utilizes cloud infrastructure to deliver virtual screening to clients
who either don’t desire to or cannot afford to maintain their
own infrastructure
Highly efficient system for managing job queuing and
maximizing the efficient use of computational resources allows
us to provide reduced-cost access to our tools for academic and
government researchers
“Bucket List” Job Queueing Model
Residual processing power in cloud
Need for low-to-no cost solutions for
academic researchers
Solution: Bucket List model for
job queueing allows unassigned
agents to perform lower-priority
jobs after finishing a paid job
and before “death” at the end
of the provisioned hour
Our Goals in Working with Chemaxon
Integration of additional chemical libraries and library filtering
tools to focus search space prior to docking
Enhancement of end-user ability to evaluate results through
integration of data analysis and visualization tools
Integration of additional licensed, proprietary, and public
domain tools
Outcome:
Products for Three Stages of Drug Discovery
Lead Generation
5 years
Candidate Identification
LeadIdentification
TargetValidation
TargetIdentification
PreclinicalCandidate
Identification PhI / IIa PhIIb PhIII Register
Lead Optimization
3 years
Product Realization
4.5 years
Fingerprinting
Modeling/Docking
Repurposing
Front-End Requirements
Spreadsheet-like viewing of compounds
Item Description Source Status
A. Structure identifier. MySQL autogenerated B. Vina Binding energy for mode 0
structure. S3 Project
generated C. Ligand Efficiency -calculated data
from number of heavy atoms in ligand and (B.)
MySQL
calculated
D. Physico-chemical properties calculated by JChem
database precalculated
E. Chemical structure (MarvinView?) MySQL calculated
Visualization Requirements
All numerical fields sortable
Data export as a spreadsheet
Mechanism for 3D view of docked structure
Drill-down for ordering requirements (integration with vendors)
3D Visualization Requirements
Using Jmol
Typical Visualizations
contact connect for 1m14, showing hydrogen
bonds between amino acid residues of a single
chain.
Why We’re Working With Chemaxon
Integration of Marvin and search tools in the web front-end
Consistent nomenclature of all library items
Automated processing of libraries
Future Goals
Build integrated suite of tools (including Zorilla applications)
Improve ancestral protein prediction in phylogenetic analysis
Answer fundamental evolutionary questions relating to
structure/function
For Further Information, contact: [email protected]
Acknowledgments
The Wyckoff Lab
Lee Likins, Scott Foy, Ming Yang
Ada Solidar (B-tech Consulting)
HaRo Pharmaceuticals
Tomasz Skorski (Temple University)
The Miziorko Lab (UMKC)
John VanNice
Andrew Skaff
Jeff Murphy (Nickel City Software)
Brian Geisbrecht (K-State)
And his lab
John Walker (SLU)
NIH 1 R41 GM 088922-01A1
NIH 2 R44 GM097902-02A1
NIH 1 R21 AI113552-01
VaSSA Informatics, LLC for major funding
Digital Sandbox KC
Missouri Technology Corporation
UMKC SBS, UMRB, UMKC FRG, KCALSI for additional funding