500,000 fish phenotypes: The new informatics landscape for evolutionary and developmental biology of the vertebrate skeleton

Paula Mabee, James P. Balhoff, Wasila Dahdul, Hilmar Lapp, Peter Midford, Todd Vision, Monte WesterfieldJournal of Applied Ichthyology 2012 · DOI

TL;DR

This paper introduces Phenoscape, a novel informatics platform that integrates vast amounts of fish skeletal morphology data with zebrafish gene-phenotype information. By leveraging ontologies, Phenoscape enables researchers to generate testable hypotheses about the genetic basis of evolutionary changes in fish skeletons, bridging the gap between comparative anatomy and developmental genetics.

What Problem Does This Paper Address?

Evolutionary developmental biology (evo-devo) faces a significant challenge in connecting the dots between the evolutionary diversity of phenotypes across species and the underlying genetic mechanisms. Specifically, two major hurdles exist:

  1. Candidate Gene Discovery: Evolutionary biologists studying the skeletal diversity of various fish species struggle to identify specific genes responsible for observed phenotypic variations. While model organisms like zebrafish have extensive genetic data, linking these genes to specific evolutionary traits in other fish is difficult.
  2. Taxonomic Distribution of Features: Developmental geneticists studying model organisms find it challenging to determine how a particular genetic mutation or developmental pathway's phenotypic effect is distributed across the vast diversity of fish species.

The sheer volume of free-text descriptions of skeletal morphology in the scientific literature makes it incredibly difficult to aggregate, compare, and analyze this data systematically. This lack of computable, integrated data hinders the generation of testable hypotheses about how evolutionary changes in development lead to phenotypic diversity.

Methods

The core of Phenoscape's approach lies in its informatics framework, which combines ontologies with a structured data representation:

  • Community Anatomy Ontology for Fishes: The researchers developed and utilized a standardized anatomical ontology for fish. This ontology provides a hierarchical and logically defined set of terms for skeletal structures, allowing for consistent representation and aggregation of morphological data.
  • Ontology-Based Methods for Free-Text Representation: They developed methods to convert complex, free-text descriptions of species' skeletal characters into a computable format. This involves annotating these descriptions with terms from the fish anatomy ontology and a quality ontology (e.g., shape, size).
  • Entity-Quality (EQ) Formalism: Phenoscape adopts the Entity-Quality (EQ) syntax, a standard in evo-devo informatics. This syntax decomposes phenotype statements into three components: the phenotypic quality (e.g., "flattened"), the entity bearing the quality (e.g., "infraorbital bone"), and the organismal entity exhibiting the phenotype. This structured representation allows for machine processing and reasoning.
  • Phenex Annotation Software: A proof-of-concept annotation software, Phenex, was developed to facilitate the manual curation of data from scientific literature into the EQ formalism using the defined ontologies.

Datasets

The Phenoscape Knowledgebase (KB) is populated with the following datasets:

  • Comparative Morphological Data: Over 500,000 taxon phenotype annotations derived from the literature on more than 2,500 teleost fish species, primarily from the Ostariophysi group. This data represents evolutionary phenotypes described as characters and character states.
  • Zebrafish Gene-Phenotype Data: Phenotype data from the zebrafish gene database (ZFIN), comprising approximately 26,934 phenotype annotations for 4,307 zebrafish genes.

These datasets are integrated within the Phenoscape KB, allowing for cross-referencing and hypothesis generation.

Key Insights and Ideas

  • Bridging the Gap: The most significant insight is the successful creation of a system that bridges the gap between evolutionary morphology and developmental genetics. By making morphological data computable and linking it to genetic data, Phenoscape provides a powerful tool for evo-devo research.
  • Hypothesis Generation Engine: Phenoscape acts as a hypothesis generation engine. By querying the integrated database, researchers can identify candidate genes for specific evolutionary phenotypes or trace the distribution of a trait across taxa.
  • Power of Ontologies: The paper strongly emphasizes the critical role of ontologies in standardizing terminology, enabling semantic integration of disparate data sources, and facilitating complex queries that would be impossible with free text alone.
  • "Translational Biodiversity": The authors propose this concept, extending the idea of "translational medicine" to biodiversity research. It highlights the potential to leverage conserved developmental genes and networks from model organisms to understand evolutionary changes across a wide range of species.
  • Data Mining Potential: The ability to browse, sort, and aggregate morphological data in novel ways opens up unprecedented possibilities for data mining and discovery in fish skeletal biology.

Key Results

  • Phenoscape Knowledgebase: The creation of the Phenoscape KB, containing over 500,000 taxon phenotype annotations for teleost fishes and integrating them with zebrafish gene-phenotype data.
  • Testable Hypotheses: The system successfully generates testable hypotheses. For example, queries can identify candidate genes for phenotypes like "gill rakers, absent" or "caudal fin, absent," linking them to specific zebrafish genes.
  • Data Integration Scale: The KB integrates data from over 2,500 teleost fish species and thousands of zebrafish genes, demonstrating the scalability of the ontology-based approach.
  • Illustrative Examples: The paper provides nine detailed examples of how the Phenoscape KB can be used to generate hypotheses for specific skeletal phenotypes, showcasing the practical application of the system.

Limitations and Future Work

  • Limitations:
    • Bias in Data: The authors acknowledge potential biases in candidate gene discovery, such as a preference for well-studied pleiotropic genes (e.g., eda, edar).
  • Missing Data: Incomplete phenotype data for certain taxa due to lack of survey or entry into the KB limits the completeness of analyses.
  • Coarseness of Annotations: The granularity of phenotype annotations for both genes and taxa can be a limitation.
  • Differences in Study Focus: Discrepancies in the focus of phenotypic study (e.g., neural system in zebrafish vs. skeletal system in comparative studies) and developmental stages can introduce biases.
  • Future Work:
    • Automating Curation: Efforts are underway to partially automate the biocuration workflow to increase efficiency.
  • Expanding Ontologies: Continued development and refinement of ontologies for broader coverage and deeper semantic representation.
  • Broader Species Coverage: Expanding the KB to include a wider range of fish species and other vertebrate taxa.
  • Developing Computational Tools: Further development of computational tools for data synthesis, visualization, and hypothesis testing.
  • Enabling Synthetic Research: The ultimate goal is to create a platform that facilitates new and synthetic research by making complex data accessible to a broad range of researchers.

Summary updated

Abstract

The rich phenotypic diversity that characterizes the vertebrate skeleton results from evolutionary changes in regulation of genes that drive development. Although relatively little is known about the genes that underlie the skeletal variation among fish species, significant knowledge of genetics and development is available for zebrafish. Because developmental processes are highly conserved, this knowledge can be leveraged for understanding the evolution of skeletal diversity. We developed the Phenoscape Knowledgebase (KB; http://kb.phenoscape.org) to yield testable hypotheses of candidate genes involved in skeletal evolution. We developed a community anatomy ontology for fishes and ontology-based methods to represent complex free-text character descriptions of species in a computable format. With these tools, we populated the KB with comparative morphological data from the literature on over 2,500 teleost fishes (mainly Ostariophysi) resulting in over 500,000 taxon phenotype annotations. The KB integrates these data with similarly structured phenotype data from zebrafish genes (http://zfin.org). Using ontology-based reasoning, candidate genes can be inferred for the phenotypes that vary across taxa, thereby uniting genetic and phenotypic data to formulate evo-devo hypotheses. The morphological data in the KB can be browsed, sorted, and aggregated in ways that provide unprecedented possibilities for data mining and discovery.
BibTeX
@article{mabee2012fish,
  author = {Paula Mabee and James P. Balhoff and Wasila Dahdul and Hilmar Lapp and Peter Midford and Todd Vision and Monte Westerfield},
  title = {{500,000 fish phenotypes: The new informatics landscape for evolutionary and developmental biology of the vertebrate skeleton}},
  journal = {Journal of Applied Ichthyology},
  year = {2012},
  doi = {10.1111/j.1439-0426.2012.01985.x},
}
Figure