caBIG GoMiner Pilot: Progress Report and caBio Integration Demo

1
caBIG GoMiner Pilot Progress Report and caBio
Integration Demo

• Rob Sfeir
• David Kane

2
GoMiner Concept of Operations

• GoMiner is a tool that leverages the Gene
  Ontology to analyze lists of genes
• Typically from a gene expression microarrays
• GoMiner classifies the genes into biologically
  coherent categories and assesses these
  categories.
• GoMiner has several major input sources
• From the GO Consortium
• Gene Ontology
• Gene Associations
• From the user
• List of Total Genes in the experiment
• List of Changed Genes in the experiment
• The insights gained through GoMiner can generate
  hypotheses to guide additional research.

Ontology
Gene Mappings
TotalFile
ChangedFile
GoMiner
3
GoMiner caBIG Pilot

• Bring GoMiner to Silver-level caBIG Compatibility
• Semantics
• UML
• Establish interoperability with other caBIG
  resources
• caBio
• caArray
• caDSR
• Web Services

4
GoMiner API UML Model

• Developed as part of the semantic mapping
  activity
• Describes GoMiners interfaces
• We have not finalized all of our modeling
  decisions
• Include GO concepts?
• We just received our EVS report
• Semantic Connector
• Manual Curation

5
caBIO BioCarta

• Leverage BioCarta to provide a biological context
  for GoMiner results
• For GO categories, provide a display that lists
  all of the BioCarta pathway maps that contain at
  least one changed gene from that category
• Include summary data so users can examine the
  most relevant pathway maps
• Use as an example of interoperability with caBIO
• Leverage the caBIO web services API

Note Builds on simpler linkouts already in
GoMiner
6
About the Plumbing

• GoMiner compiles a list of genes
• Opens a web services connection
• For each gene, GoMiner issues a web services
  request to find all of the pathway maps
  associated with that gene
• The results are tabulated and presented to the
  user
• Remove duplicates
• Count changed genes on each pathway map
• A plain browser-based web page request is used to
  display user-selected maps

7
The Demo
8
For the Propeller Heads Code Details

• Easiest way to get going is to
• Look at their code samples, you have everything
  you need there. They have an excellent test
  class to run all sorts of selects.
• Base your code on the test cases.
• There is a webservices directory in the
  distribution, and everything you need is in
  there. No need to compile or generate anything.
• When making queries to pathways, genes, taxons
  etc remember to register the Gene type along
  with the other types you need to get back

final QName qnGene new QName("urnimpl.domain.ca
bio.nci.nih.gov""GeneImpl") call.registerTypeMap
ping(GeneImpl.class, qnGene, new
BeanSerializerFactory(GeneImpl.class, qnGene),
new BeanDeserializerFactory(GeneImpl.class,
qnGene)) //This is added to the second argument
of the //Call object to search by
gene call.addParameter(arg2, qnGene,
ParameterMode.IN)
9
More Code Details

• Results of the webservice call returns a Object
  object, which contains the objects you were
  selecting. In our case PathwayImpl.
• Depending on what youre doing, the Object may
  need to be rebuilt to an ArrayList.
• In our case, to get things displayed properly in
  a JTable, we needed to rebuild the array so we
  could gain a Key Value pair for the TableModel.

10
Observations

• The version of caBio in the 3.0 caCore release
  has much stronger web services features than
  earlier caBio releases
• Faster responses
• Easier to get to the data of interest
• Described using WSDL
• This is not an ideal interoperability
  demonstration
• Our implementation uses client library that was
  included in the caCore distribution.

11
Open Issues

• All of the following are issues that we
  encountered, but not yet resolved at this point
• The pathway query API includes a collection to
  list the genes on the pathway
• Put for our queries, this collection always null
• We are only getting human pathways back
• We are unsure why we are not getting mouse
  pathways which appear on the NCICB site
• We have been unable generate stubs from WSDL.
• Workaround is to simply use the client library
  included in the caCore distribution