Advances in the biomedical applications of the EELA Project

1
Advances in the biomedical applications ofthe
EELA Project

• Rafael Mayo
• CIEMAT
• HealthGrid Conference
• Geneva, 24-27.04.2007

2
Some of the slides have been taken fromprevious
presentations made byVicente Hernández
(UPV)(EELA Biomed Task Leader)
3
Contents

• The EELA Project
• Biomedical Applications
• BiG (Blast in Grids)
• MrBayes
• GATE
• WISDOM (Wide In Silico Docking Of Malaria)
• Summary of Achievements
• Future Plans

4
BUILD A BRIDGE BETWEEN CONSOLIDATED
E-INFRASTRUCTURE INITIATIVES IN EUROPE AND
EMERGING ONES IN LATIN AMERICAREINFORCE
COLLABORATION BETWEEN LATIN AMERICA AND EUROPE
AIMS
5
PARTNERS
Italy INFN Portugal LIP Spain CIEMAT, CSIC,
Red.ES, UC, UPV CERN
Argentina UNLP Brazil CEDERJ, RNP, UFF,
UFRJ ChileREUNA, UDEC, UTFSM CubaCUBAENERGIA Me
xicoUNAM PeruSENAMHI VenezuelaULA CLARA

• 10 Countries
• 21 Partners

6

MANAGEMENT STRUCTURE

• 4 WORK PACKAGES DIVIDED IN TASKS
• WP1 PROJECT ADMINISTRATIVE AND TECHNICAL
  MANAGEMENT
• WP2 PILOT TESTBED AND SUPPORT
• Task 2.1 Coordination of e-Infrastructure
• Task 2.2 Certification Authorities and Virtual
  Organizations
• Task 2.3 Pilot Testbed Operations
• Task 2.4 Network Support and Operation
• WP3 IDENTIFICATION AND SUPPORT OF GRID ENHANCED
  APPLICATIONS
• Task 3.1 Biomedical Applications
• Task 3.2 High Energy Physics Applications
• Task 3.3 Additional Applications (Education and
  Climate in the Grid Environment)
• WP4 DISSEMINATION ACTIVITIES
• Task 4.1 Dissemination
• Task 4.2 Knowledge Dissemination

7
THE EELA WEB SITE
http//www.eu-eela.org
8
THE EELA e-INFRASTRUCTURE
9
WORK PACKAGE 2 STATUS

• 734 CPUs
• 61.2 TB

10
EELA CAs
11
THE EELA PILOT TEST-BED
RC Resource Center T-I t-Infrastructure Center
12
WORK PACKAGE 4 STATUS

• 15 TUTORIALS More than 1200 participantsdays
  delivered so far in 2006!
• For grid users (total of participants 300)
• Madrid - Spain (20-22 February 2006)
• Merida - Venezuela (27-29 April 2006)
• Itacuruçá - Brazil (26-28 June 2006)
• Mexico City - Mexico (28-30 August 2006)
• Santiago de Chile (6-7 September 2006)
• Madrid Spain (16-18 October 2006)
• Merida Spain (7-9 November 2006)
• La Plata Argentina (11-12 December 2006)
• Bogota Colombia (6-7 March 2007)
• For grid system administrators (total of
  participants 160)
• Madrid - Spain (20-24 February 2006)
• Itacuruçá - Brazil (26-30 June 2006)
• Mexico City - Mexico (28 August - 01 September
  2006)
• Madrid Spain (19-20 October 2006)

13
WORK PACKAGE 4 STATUS

• 1st EELA GRID SCHOOL. Itacuruçá - Brazil (4 -15
  December 2006)
• Training for gridification of new applications
  to be run on EELA infrastructure (23
  participants) 7/8 applications gridified
• Climate applications, proposed by UC-SP,
  SENAMHI-PE and UDEC-CL (EELA applications)
• Dist Simulation of Multiple Events, proposed by
  UNESP-BR, UNICAMP-BR and USP-BR (non-EELA
  application)
• EMBOSS, proposed by UNAM-MX (non-EELA
  application)
• LEMDist, proposed by UNAM-MX (EELA application)
• LMS in Grid Env, EELA application proposed by
  CITMATEL-CU (non-EELA partner) and CUBAENERGIA-CU
  (EELA partner)
• SATyrus, proposed by UFRJ-BR (EELA application)
• SegHidro, proposed by UFCG-BR (non-EELA
  application)
• VoD, proposed by UFRJ-BR and CEDERJ-BR (EELA
  application)

14
WORK PACKAGE 4 STATUS

• DISSEMINATION 2 bulletins, 2 brochures, 4
  posters, 17 press releases, 10 Inf. Sheets, 72
  news

15
WORK PACKAGE 4 STATUS

• 5 WORKSHOPS
• 1 CONFERENCE
• Future events
• 2 WORKSHOPS (Cuba Mexico)
• 1 CONFERENCE (with BELIEF Brazil)
• Potential new EELA collaborators or partners 51

16
WORK PACKAGE 4 STATUS

• Survey of Communities results
• 46 LA communities
• 5 European communities from France,
  Luxembourg, Spain, UK

Brazil
Chile
Peru
Mexico
EU
Colombia
Ecuador
Nicaragua
Venezuela
El Salvador
17
Biomedical Applications
WORK PACKAGE 3
18
Status of the Applications

• EELA Applications
• New Applications
• BiG (BLAST in Grid) is a Grid-enabled BLAST
  Interface
• BLAST (Basic Local Alignment Search Tool) is a
  Bioinformatics Procedure Applied to Identify
  Compatible Protein and Nucleotids Sequences in
  Protein and DNA Databases.
• MrBayes is a Tool for Phylogeny Studies
• A Phylogeny is a Reconstruction of the
  Evolutionary History of a Group of Organisms.
• EGEE-Ported Applications
• GATE is an Environment for the Monte-Carlo
  Simulation of Particle Physics Emission in the
  Medical Field.
• It is Focused Towards Thyroid Cancer and
  Treatment of Metastasis with P32.
• WISDOM (Wide In-Silico Docking Of Malaria) is a
  Deployment of a High-Throughput Virtual Screening
  Platform in the Perspective of In-Silico Drug
  Discovery for Neglected Diseases.

19

• GATE
• Geant4 Application for Emission Tomography

20
GATE

• Gate is Installed on
• tochtli.nucleares.unam.mx GATE-1.0.0-3
• ramses.dsic.upv.es GATE-1.0.0-3
• grid012.ct.infn.it GATE-1.0.0-3 and
  VO-biomed-GATE-2.2.0-3
• ce-eela.ciemat.es VO-biomed-GATE-2.1.0-1
• ce01.eela.if.ufrj.br
• ce-eela.ciemat.es
• A demonstration was performed in the frame of
  the EU-LAC Summit, Lisbon 28-29 April 2006

21
GATE

• User Community
• It is focused towards two main oncological
  problems
• Thyroid cancer.
• Treatment of metastasis with P32.
• 9 centers in Cuba are interested
• Due to the Lack of Bandwidth from/to Cuba a Local
  Configuration is Being Set-up.
• A dark fibre is settled to Venezuela

22

• WISDOM
• Wide in Silico Docking on Malaria

23
First EELA WISDOM Data Challenge

• Objective and Means
• Docking of Two Targets of a Segment of the
  Plasmodium vivax Genoma With Respect to the
  Non-Redundant Ligand Database.
• Execute a First Test on the Infrastructure.
• Using the WISDOM Scripts Developed in EGEE.
• UPV coordinated
• Outcome
• Last Job of the First Target Finished on January
  the 1st, 2007.
• First Data Challenge Experiment on Reaching 2421
  from 2422 Jobs.
• 53 Gbytes of Results Obtained.

24
First EELA WISDOM Data Challenge

• Problems Faced and Lessons Learnt
• Difference of Scale Between EGEE and EELA. The
  Scripts were Not as Efficient for a Reduced
  Infrastructure in Which the Sites are Less
  Powerful
• Automatic Resubmission of Scheduled Jobs Make
  Throughput to be Reduced as the Number of Jobs
  Remaining Decreased.
• Automatic Resubmission Delay was Too Short
  Considering the Number of Available Resources.
  Changes on the Scripts Were Made on-the-fly.
• Two Unexpected Power Cuts, one Internet Cut.
• A New Status was Created to Restart in a
  Controlled Way the System.

25
First EELA WISDOM Data Challenge

• Problems Faced and Lessons Learnt
• Lack of Response of BDII
• The BDII at the CIEMAT Experimented Some
  Malfunctions that Reduced the Performance in the
  RB and the Catalogue.
• Expiration of Certificates in Hosts and Users,
  Change on the Certification Authority.
• Launching User Certificate Expired, and Change on
  the CA Subject Produced Several Problems. The
  Scripts Needed to be Re-Launched.
• Several Hosts Expired the Certificates and
  Required to Pause the Process During some Hours.
• The CRLs of Spanish Sites were Outdated for Two
  Days.
• Corruption of Files in SEs
• During the Copying Process, Several Fragment of
  the ZINC Database were Unsuccessfully Copied.
• One File is Corrupted in the Original
  Repository!!
• No one has Reached the 100 of Completion so Far.

26
First EELA WISDOM Data Challenge

• Lessons Learnt for New DCs
• Scripts Must be Adapted to Deal with the
  Conditions of EELA Infrastructure.
• Double Check the Input Files, even if it Takes
  Several Days.
• Check the Evolution of the Process Several Times
  Per Day.
• Check Wall-Time in All Resources.
• Unavoidable Problems
• Unexpected Power Cuts or Internet Cuts.
• Other Site-Specific Administration Problems.
• All These Problems (Normal in a First Large-Scale
  Execution) were Solved with the Strong Support
  of the Infrastructure Managers of EELA and the
  Team of Vincent Breton.

27
First EELA WISDOM Data Challenge

• Statistics
• Starting Date October the 23rd, 2006.
• Ending Date January the 1st, 2007.
• Number of Original Jobs of the First Target
  2422.
• Number of Jobs Successfully Completed 2421.
• Total Number of Submitted Jobs 109551.
• Most of the Submitted Jobs Did Not Reached the
  Execution Queues.
• Average Computing Time per Job 2065 Minutes.
• Total Effective Running Time 228 cpu/days.
• About Half of the Time the System were Not
  Producing Results. Throughput got an Important
  Decay As DC was Progressing.
• Results Obtained 53 Gbytes.

28
First EELA WISDOM Data Challenge
2500
Changes on the resubmission.
2000
CRL Problem
1500
Script does not Restart Properly
1000
CA Change
Power Cut
500
0
23/10/06
01/11/06
12/11/06
22/11/06
02/12/06
12/12/06
22/12/06
01/01/07
29
First EELA WISDOM Data Challenge

• Sites Used
• CIEMAT, Since Beginning.
• GRYCAP-UPV, Since Beginning.
• INFN , Since Beginning.
• UNAM, Since Beginning.
• UFRJ, Since Beginning.
• UNICAN, LIP and CERN During the DC Execution.
• More than 40 of the Jobs Were Effectively
  Executed in LA Sites.

30

• BiG
• BLAST in Grids

31
Blast in Grids Approach

• Use of MPI-Blast kernel
• Enhanced security through a MyProxy server
• Fault tolerant on the client and server side
• Embeddable on a stand-alone application or web
  portal
• Splitting of input sequences and reference
  databases into multiple jobs. Deals with multiple
  databases simultaneously

32
Blast in Grids Improvements on the Service

• Updates on the Grid Service
• Security Model
• In the Initial Design, The Private Key of
  Operation Users Must be Hard-Coded on the Portal
  and This Represents a Security Risk.
• The New Approach is to Use a MyProxy Repository
  To Store Long-Living Credentials and Download
  Delegated Proxies on Require.
• This Will Also Solve the Problem of Long-Living
  Jobs by using Proxy Renewal.
• Postprocessing of the Output of the Execution
• Improve the Output Format of the Processing.

33
Blast in Grids Improvements on the Security Model

• Main Problems Faced
• Uploading VOMS Credentials A myProxy Credential
  is Uploaded in a Proxy Server.
• It is Suggested that VOMS Attributes Should be
  Added After the Retrieval of a Delegated Copy of
  the Proxy gt It Does not Work.
• VOMS Attributes Could not be Uploaded With
  Standard MyProxy Commands gt Use a New Version
  from INFN.
• VOMS Credentials Duration and Proxy Renewal A
  Delegated myProxy Credential Needs to Be Renewed
  for a Long-Living Job.
• It Does Not Work with VOMS Credentials. VOMS
  Life-Time is 24 Hours gt Unsolved Problem for
  Long-living Executions.
• Other Problems
• Incorrect Configuration of Automatic Renewal on
  RBs
• Not All Resource Brokers Have these Features
  Correctly Implemented gt EELA RB is Correctly
  Working.
• Missing Documentation
• Available Recipes are Incorrect.
• There is no Official Documentation lt Important
  Assistance from Catania.

34
Blast in Grids Improvements on the Postprocessing

• Fixed Number of Returned hits MPIBLAST Bug
• MPIBLAST has a bug and it Does not Return the
  Selected Number of Hits but All of them Instead.
• This Bug has Been Fixed in the BiG Service.
• New BiG System Monitoring Methods
• Methods to Explore the Installed Databases.
• To Obtain the Input Arguments of a Query.
• A more Accurate Number of Alive Processes Status.
• New Postprocess Methods
• Four new Methods Added
• Two for Processing multiFASTA Results
• Return the Obtained Results (Total and
  Non-downloaded Results) in multiFASTA format.
• Two for Hits Summary.
• Similar to Above but Using Hits Definitions.

35
Blast in Grids Usage Report

• Period Jul06-Dec06.
• Usage Statistics
• Number of Jobs 284.
• CPU Consumed 173 CPU/Days.
• Resources Used ramses.dsic.upv.es2119/jobmanager
  -pbs-biomedg.
• BiG is Being Used at the University of Los
  Andes to Work on the Complete Genome of the
  Plasmodium falciparum for the Identification of
  DHFR Antigenic Proteins.

36

• MrBayes
• Bayesian Inference of Phylogeny

37
Phylogeny

• A Phylogeny is a Reconstruction of the
  Evolutionary History of a Group of Organisms.
• Most Approaches Depend Upon a Mathematical Model
  Describing the Evolution of Characters Observed
  in the Species Included, and are Usually Used for
  Molecular Phylogeny Where the Characters are
  Aligned Nucleotide or Amino Acid Sequences.
• Proposal Use MrBayes
• MrBayes is a Widely Used Bayesian Inference
  Application for Phylogeny.
• It is MPI-Enabled.
• Consumes Much Memory and Computing Time.

38
Phylogeny

• A Grid-Service to Run Parallel MrBayes Executions
  is Being Implemented.
• It Requires MrBayes to be Installed at the Sites
  and Hides the Parallel Configuration for the
  Execution.
• The Structure is Similar to the BiG Service, Even
  Simplified.
• Portal Interface in Progress.

Interface
39
WP3 web pagehttp//www.eu-eela.org/eela_wp3.php
WP3 documentshttp//documents.eu-eela.orgCont
actsVicente Hernández vhernand_at_dsic.upv.esRafa
el Mayo rafael.mayo_at_ciemat.es
40
Thanks for your attention