Phenoscape Grant Renewal Workshop/Notes

Phenoscape II: Tuesday, April 28, 1:00-5:30

Participating: Phenoscape: Paula Mabee, Todd Vision, Monte Westerfield, Hilmar Lapp, Wasila Dahdul, Jim Balhoff, Cartik Kothari, Peter Midford, John Lundberg; Suzanna Lewis, Judy Blake, Peter Vize, Anne Maglia

• Will be extending the taxonomic scope to extant and extinct vertebrates
  • Create an annotation database spanning the taxonomic scope and their anatomy ontologies
  • Three multi-species anatomy ontologies are being developed: teleosts, amphibians, mammalian anatomy
  • Tree mapping: show when an evolutionary phenotype first appeared
• EQ support in all involved model organism databases
  • Structure of MP is not consistent with anatomy, or with PATO
  • Mapping from MP to EQ syntax is being worked on (George Gkoutos)
    • How is this supported in MGI?
      • Could spit back EQ, but could not allow users to search
  • MGI isn't in a position to use EQ internally, but for example Phenoscape II could provide an EQ-view on mouse phenotypes, using the decomposition of MP into anatomy (or entity) and quality term cross-product
  • Full EQ annotation for OMIM is being planned but not yet funded
  • MGI, Xenbase, and ZFIN all have links from their phenotype data to OMIM
• Phenoscape II: Development of anatomy - new queries supported; would generate testable hypotheses; how exactly do those queries rely on developmental data?
  • Different approaches between MGI and ZFIN:
    • MGI uses complete anatomy at different developmental stages
    • ZFIN uses one anatomy ontology and adds start and end dates to indicate the developmental period during which it appears (adult structures don't have an end date)
    • Xenbase uses the ZFIN approach
  • Candidate genes for evolutionary change could be derived from anatomy-annotated gene expression studies during development; Where do mouse and Xenopus share gene sets in early development?
    • genes responsible for or associated with morphological changes during individual development could be candidates for evolutionary change
    • evolutionary phenotype changes could be used to query morphological changes during development
  • Can we enable queries to generate hypotheses based on ontogeny reflecting evolutionary history?
  • Powerful tool in interpreting microarray data downstream from node; know candidate genes without mutants using this approach
  • See phenotypic differences between 2 taxa – where do you see similar changes?
  • Cell types and associate with time and space – can reason from, neural crest, etc.

• • • Evolutionary developmental data could come from medaka, stickleback
• Sequence of steps:
  1. Ontology building
     • Adult phenotype annotations for mutants in Xenbase
     • Presently only 3 species for amphibians: Xenopus, Dermophis, Salamandria
     • Development decoupling from anatomy ontology in mouse
     • Expanding mammalian anatomy to include extinct species
     • Anatomy for amniotes (the clade including mouse and dinosaurs), including extinct taxa (such as dinosaurs)
       • Scope of this could be overwhelming
       • Should be strongly driven (or staged) according to the character matrices to be annotated
       • Chicken anatomy could provide a starting point?
       • There is work on bird anatomy that uses a latin naming scheme
       • Work on any smaller-scope anatomy (or multi-species anatomy) ontology will contribute terms that apply more broadly
       • There may not be very many neomorphs between birds and mammals, though there are a few areas such as the digits in birds where the exact homology relationships aren't as clearly agreed upon.
  2. Mammalian phenotype decomposition
     • Need support for QC'ing the results of automated decomposition
     • Alignment of MP with mouse anatomy
  3. Annotation
     • Which published character matrices are there for dinosaurs?
• Ontology mapping and alignment
  • Need to align teleost and amphibian anatomy to mouse.
  • Mouse and chicken need to be mapped to amniote anatomy. Mouse is a better start because there are genetic data.
• Reconciling character-derived trees and character definition and use between different matrices and trees that share taxa (Paul)
  • TaxonSearch
  • Formalizing and generating grammar of character coding and character state definition
  • Analyzing character usage (shared, not shared, rejected) between trees that share taxa
  • Conflicting phylogenies for the same set of taxa often turn out to be based on character codings that are not consistent or not compatible

Wednesday morning

• Nomenclatural history of terms in ontologies
  • Current exchange format standards don't support this really beyond obsoletion
  • ZFIN tracks within the database the complete nomenclatural history of gene names
  • RDBOM tracks literature and author attribution for anatomy terms but not yet nomenclatural history
• Character quality profiles for data matrices
  • Addresses the question of "Where in the skeleton are the changes occurring that drive the phylogeny"
  • Distribution of characters across the skeletal anatomy, at different levels of the hierarchy
  • Distribution of missing data across the anatomy, at different hierarchy levels
  • How many taxa have how much missing data, distributed on which anatomical parts
• Use case: Character redundancy
  • Deriving a character profile across the anatomy could help visualize the redundancy between annotations
• Use case: Mapping the evolutionary characters that lead (and distinguish) to a model organism. Subsequently, see whether these evolutionary changes and the order in which they appear to the ontogeny of the organism.
  • Will need to deal with character gaps and with character redundancy for this.

Specific goals:

1. Expand taxonomic coverage of Phenoscape to Vertebrata
   • Mouse is done but needs to be rearranged (MP decomposition)
   • Amniotes need to be done from the ground up

Wednesday afternoon

• Problem of lossy transformation of legacy character data
  • EQ annotation is often at a lesser granularity (or level of detail) than the original free text description
  • This isn't due to any technological obstacles but rather due to the limited resources available for annotation.
  • Annotation effort is best focused to a granularity level where it suffices to solve a use-case of interest.
  • Annotating at very granular levels is entirely possible but can take a lot of time because many of the ontology terms needed are likely not to be present yet in the ontology.
  • How does this impact phylogeny and data matrix reconciliation, or phylogenetic reconstruction on the basis of the EQ annotations?
• Use-case for development
  • Heterochrony would be interesting if one can pull it out of the database
  • Developmental ordering relationships currently used are confined to develops_from and transforms-into.
    • This is missing yet for mouse.
    • Wouldn't really allow inferring heterochrony.
    • Would allow though to place and compare the placement of phenotypes into a developmental chain of transformations.
• Demonstration of Phenoscape KB
  • Some of the features we are developing (the 3 principle queries) would be very useful to MOD users
• Informatics goals for Phenoscape 2:
  • Supporting units for measurements
  • Allowing others to compare their phenotype data to the knowledge base
  • Generic ontological queries, such as a SPARQL endpoint
  • Private data overlay
  • Linked Data and semweb integration
  • Tree visualization
  • Triple store technology evaluation
  • Species ID
  • Phylogenetically informative distance metric based on EQ assertions
  • NLP-based text processing, Mass curation

Idea bin

1. Intermine connection (multiple model organisms, AJAX-based widget for displaying protein family tree)
2. Some phenotypes imply developmental abnormalities, differences, or variation.
   • For example, a "poorly ossified cranium" in an adult amphibian implies that the developmental process was delayed or did not complete.
3. How does logical inference of homology propagate over develops-from relationships.
   • E.g., if A and B are asserted homologous, and C develops from A and D develops from B, are then C and D inferred as homologous, and are C and B inferred as homologous.