Needs Analysis Workshop/Summary

From phenoscape
Revision as of 19:42, 18 January 2008 by Jpb15 (talk | contribs) (Data model)

Background

The PhenoScape Needs Analysis Workshop was held September 17-18, 2007 at NESCent. We invited 9 scientists from the fields of morphology, development, evolution, genetics, and ichthyology to help us define what informatic tools need to exist in order to enable synthetic research that takes full advantage of the accumulated data in each of these fields. We hoped to identify driving biological questions and use cases to guide the development of our software tools and database.

Agenda

After being introduced to the project goals, the meeting participants each provided a brief chalk talk outlining their integrative research problems. The participants then split into two break-out groups, each arranged around either the developmental genetics of morphology or the evolution of morphology. At the end of the day the groups merged again to discuss the research questions arising in the breakout groups.

The second day began with three break-out groups: correlations, phylogeny, and semantics. As before, the whole group then discussed the uses cases arising in the break-out groups, and tried to identify priorities for the project.

Driving questions/use cases

The following is a bullet-point summary highlighting both research questions and software use cases produced by the discussion. Minutes from the relevant session are linked next to each point.

  • Use the database to identify discrete evolutionary modules BG1
    • Depends on identifying patterns of evolutionary covariation among phenotypes
  • For a branch hypothesized to be under positive selection (from e.g. a molecular analysis), identify candidate phenotypes that could have been selected for BG1
    • Requires mapping phenotypic changes on phylogeny, visualizing synapomorphies
    • Topology can be modified and the results of modification visualized
  • Use the database to help with negative information - what knowledge is missing that is required to answer a particular question? BG1
  • Integrate data regarding ecologic conditions, life history, adaptation, and evolutionary time with phenotypic and genetic data to elucidate mechanisms of phenotypic change BG2
    • must integrate with existing databases, literature, across biological levels
    • use other factors (environment, ecology) as constraints on data presentation
  • User adds own annotations using modified curator interface, compares own data with database BG2, PB
    • user may also want to store media such as images and movies
  • Enrichment analysis: given a list of genes, find which phenotype terms are overrepresented among them RO1
  • User is notified of additions or changes to the database relevant to him/her RO1
    • save user-defined queries as RSS feeds
  • Capture and represent within-species variation in phenotypes and genotypes RO1
    • describe polyphenism dependent on environmental factors or variable gene expression
    • view population frequencies of polymorphic phenotypes
    • describe phenotypic change through developmental time series
  • View traits related to particular parts of anatomy - just tail, or just vertebra CB
  • User view ancestral character states on tree using default reconstruction methods CB
  • User downloads data in a form useful for applying their own methods of character reconstruction CB
  • Find cloud of similar annotations using summary matrices CB
    • traits vs traits, traits vs genes, genes vs genes
    • cells would be number of branches on which the two traits change in common (traits vs traits), or number of mutants they have in common (traits vs traits, and traits vs genes), or number of traits in common (genes vs genes)
  • User uses interactive tree view to select taxa for summary matrix CB
  • User selects one or more regions on body plan view to view traits(genes) affecting all CB
    • then map corresponding traits on tree
  • User creates and edits a tree using MacClade-like interface, then visualizes character changes on tree PB
  • User may want to work with data regarding mutations actually underlying evolutionary change SB
    • different genetic states in different species

Requirements arising from the use cases

Overview

The use cases above present requirements for both our scientific framework/data model and our software tools (curation tools and web application). A good overview can also be found here.

  • Provide character-state reconstructions on given phylogenies
    • analysis
    • visualization
  • View and interactively/graphically edit phylogenies containing some subset of the taxa in the database
  • Connect with and integrate data from external databases
    • ZFIN, GenBank, other genomics-oriented databases
    • Ecological, geographic, historical data - perhaps through museum catalog data
  • Provide curation interface, storage, and analysis integration for private user data
  • Save user queries as RSS feeds
  • Filter view of data by ontology terms: anatomy, species, qualities
    • Provide advanced filtering functionality by means of ontology structure
  • Provide views of data in downloadable forms convenient for further analysis
    • simple tables
    • generated NEXUS format
  • Provide summary matrix views showing covariance/co-occurrence of various datatypes
    • Should be easily linked to other types of views (tree, etc.) restricted to data in particular cells
  • Provide body plan view linked to anatomical terms for selection and filtering
  • Support genetic diversity data along with phenotypic diversity data - molecular changes directly related to evolutionary changes

Web application

The requirements primarily address capabilities of the web application used as a front end to our data collection. In many cases they do not significantly diverge from capabilities we would already be planning - using filters to restrict the data being fed into a view, providing link-outs to other genomic or museum databases, integrating data from those databases in useful ways, providing a means for a user to save and check favorite queries, and exporting the data to their local computer.

Some of the proposed views are more challenging but doable. The phylogenetic tree view is not a new idea, but the participants desired some advanced capabilities from it. Displaying character state changes on the tree requires programming the graphics to do that, as well as computing the reconstruction of ancestral states. Because there are several ways to do that reconstruction, we might choose a basic method and let users download the data for more complex analyses. All methods of reconstruction rely on the data being in character matrix form - more on that below. Further capabilities desired of the tree view lead to an almost complete implementation of the MacClade/Mesquite tree editing features within a webpage (presumably via JavaScript). This would be a significant project, although probably quite useful to the community. Some beginnings exist in other projects on the web.

The body plan view will require annotation of a complex image with various ontology terms. This could be fairly coarse, or as complex and fine-grained as we want to make it.

Data model

Matrices displaying covariation/co-occurrence of phenotypic traits with other traits (or other data) will in many cases require ancestral state reconstruction, already mentioned as a desirable analytic view of the data. This requires a tree, a reconstruction algorithm, and data in the proper format. While our data describe character diversity, the annotations require processing to be converted into evolutionary character matrix form. This processing relies on distinguishing between attributes and values in the PATO ontology and following a particular algorithm. Since this process is central to many of the forms of data analysis that came up in the meeting, I think we need to look more carefully at how we want matrix generation to work and what is possible with our annotation strategy and ontologies. Some of this is discussed in EQ for character matrices.