D2OL and Community TSC

1
D2OL and Community TSC

• Informatics Staff
• Wolfgang Hinz
• Jian-Quan Chen
• Ken Weilbacher
• Daniel Yu

2
Presentation Outline

• Introduction to D2OL
• AutoDock 3.0
• Community TSC
• D2OL SARS, Bio-terrorism
• Libraries
• Results
• The Project in Numbers
• D2OL Version 2.0

3
Introduction to D2OL

• D2OL is a distributed computing platform for
  Flexible Ligand Docking on PCs

• Designed to harness underutilized computing
  power for computationally demanding work
• Allows the fast and efficient screening of
  millions of compounds against
• hundreds of targets

• Allows quick optimization of parameters based
  on large data sets

4
AutoDock 3.0

• D2OL is based on AutoDock 3.0
• Two broad categories for automated docking
• Matching Methods create a model of the active
  site, typically including sites of hydrogen
  bonding and steric accessibility then attempt to
  fit body to structure of active site by matching
  its geometry (e.g. Dock)
• Docking Simulation The ligand begins randomly
  outside the pocket and the program explores
  translations, orientations and conformations
  until an ideal state is found (e.g. AutoDock)
• Disadvantages of Docking Simulations
• Generally slower
• Advantages of Docking Simulations
• Utilize more detailed molecular mechanics to
  calculate the energy of ligand in the context of
  the putative active/binding site
• Molecular docking is a difficult optimization
  problem, requiring efficient sampling across the
  entire range of positional, orientational, and
  conformational possibilities.

5
AutoDock 3.0

• Hybrid Search Methods in AutoDock
• Genetic algorithms for global search aspects
• Local search methods for refined search aspects
• A combination of both methods is used, and is
  generally referred to as a Lamarckian genetic
  algorithm
• This method applies a Lamarckian model of
  genetics, in which environmental adaptations of
  and individuals phenotype are reverse
  transcribed into its genotype and become
  inheritable traits.
• Chromosome string of real-valued genes
• Three Cartesian coordinates for ligand
  translation
• Four variables defining a quaternion for ligand
  orientation
• One real-value for each ligand torsion obtained
  (defined by the torsion tree which is generated
  by AutoTors)
• There is a one-to-one mapping from the ligands
  state variables to the genes of the individuals
  chromosome.

6
AutoDock 3.0

• A generation consist of the following steps
• Mapping and fitness evaluation
• Selection (which individuals will reproduce)
• Crossover (performed on random members, user
  defined rates, two-point crossover with breaks
  between genes)
• Mutation (addition of a random real number with
  Cauchy distribution)
• Elitism (user defined number of top candidates
  that automatically make it into the next
  generation)

7
Community TSC

• Potential drug targets in PI3K-Akt-Tuberin
  Pathway
• 4 Targets have been identified so far
• (PTEN)
• FRAP (mTOR)
• EIF4E
• PI3K
• A fifth target is being prepared (Akt)
• The area of interest for small molecule docking
  is defined by active sites identified for each
  target.

8
D2OL SARS, Bioterrorism

• Targets originate from potential biological
  warfare agents and include
• Anthrax Lethal Factor
• Smallpox Topoisomerase
• Ebola
• Recent outbreak of SARS gave opportunity to help
  with development of antiviral agents
• First target released was TGEV (Pig Corona virus
  Main Protease)
• Second target released was Homology Structure
  derived from TGEV and Human Corona virus Main
  Protease
• Third and fourth target to be released next week
  will include Human Corona virus Main Protease and
  Rhinovirus Main Protease.

9
Libraries

• Libraries are prioritized according to
  accessibility
• Immediately accessible to TRI are
• NCI Open compound library
• ICCB Commercial Libraries
• Recipe for Library preparation
• 2D to 3D conversion
• Removal of counter ions
• Molecular weight filter (200-700)
• Removal of compounds with elements other than C,
  N, S, O, H, P, F, Cl, Br, I
• Removal of compounds that do not contain H and C
• Add partial charges using AutoDock tools
• Covert files from mol2 to pdbq (AutoTors)
• Remove compounds that have more than 16
  rotateable bonds
• Generate Docking- and Grid Parameter Files

10
Results

• Target PI3K (Community TSC)
• Candidates 22,000 NCI Compunds
• 6 Control compounds (known inhibitors)

11
Results

• The final model used in the evaluation of the
  docking energy in AutoDock 3.0 has a residual
  standard error of 9.11 kJ mol-1 (2.177 kcal mol-1)

12
The Project in Numbers

• Community TSC
• 15,000 registered users
• Daily average of 50 new registrations (peak
  125/day)
• 1,500 nodes on-line per day (peak 2400)
• 12,000 downloads in total (this figure is lower
  due to mirror sites)
• - Daily average of 75 new downloads
• 500,000 candidates processed per week (all
  targets)
• 4 Targets deployed
• NCI Library of 180,000 deployed for all targets

13
The Project in Numbers

• The distribution of Operating Systems
• (Community TSC)
• For 112,011 docking results returned
• Windows 107,930
• MacOSX 515
• Linux 3566
• Solaris 0

Percentage of Operation System
14
The Project in Numbers

• D2OL SARS/Bioterrorism
• 40,000 registered users (24,000 have registered
  since release of SARS candidates)
• 1,000 new registrations per day (peak 3,000)
• 4,000 users on-line per day (700 before SARS)
• Peak values 5, 500 (1,100 before SARS)
• 85,000 downloads (57,000 since release of SARS)
• 1,700 downloads per day (peak once at 8300 and
  twice at 6,000)
• 900,000 candidates processed per week (all
  targets)
• 5 Targets deployed
• NCI Library of 180,000 deployed for all targets

15
D2OL Version 2.0

• Completely redesigned system
• the new design is intended to remove all
  possible software related bottlenecks. The
  scalability of the project will only depend on
  the number of servers and the bandwidth available
  to the project.
• Bandwidth requirements will be further scaled
  down by preprocessing results on the client.
• Improved monitoring capabilities.
• Improved statistics.
• Results analysis system (TRI Development).
• Improved documentation of code and system
  architecture.

16
Acknowledgements

• Sengent
• Scott McFarland
• Anthony Glaviano
• Antonio Pila
• Gioel Molinari
• The Rothberg Institute
• Bonnie Rothberg
• Jonathan Rothberg
• Yale University
• Tian Xu

17
An Environment for Creativity
18
Chemistry

• Wolfgang Hinz

19
Chemistry Lead Compound

• Lead obtained from primary screen of the
  Microsource library.
• Lead confirmed by several secondary screens
  including dose response

20
Chemistry - Analogues

• Three analogues were synthesized to explore the
  activity of puromycin.

21
Chemistry - Analogues
22
Chemistry - Analogues
23
An Environment for Creativity