I attended the Evolution 2014 meeting a few months ago in Raleigh, NC, and presented a poster on Phenoscape’s curation effort: “Moving the mountain: How to transform comparative anatomy into computable anatomy?”, with coauthors A. Dececchi, N. Ibrahim, H. Lapp, and P. Mabee. In this work, we assessed the efficiency of our workflow for the curation of evolutionary phenotypes from the matrix-based phylogenetic literature. We identified the bottlenecks and areas of improvement in data preparation, phenotype annotation, and ontology development. Gains in efficiency, such as through improved community data practices and development of text-mining tools, are critical if we are to translate evolutionary phenotypes from an ever-growing literature. The poster was well received and several researchers at the meeting were interested in learning more about open source tools for phenotype annotation.
Our paper describing the Vertebrate Taxonomy Ontology (VTO) is published! See: http://www.jbiomedsem.com/content/4/1/34 .
One primary objective for Phenoscape and similar projects is to aggregate phenotypic data from multiple studies to named taxa, which in many phylogenetic studies are species but also might be at higher taxonomic levels such as genera or families. While there are many widely used taxonomies that include rich sampling of species and higher taxa, for example Bill Eschmeyer’s widely used Catalog of Fishes, there are few vetted “bridging” taxonomies that allow for aggregating data across, say, fishes, amphibians, and mammals. This problem becomes even more acute when you consider integrating data for extinct taxa as well. As a first step towards addressing this issue for vertebrates, we created the Vertebrate Taxonomy Ontology (VTO) that brings together taxonomies from NCBI, AmphibiaWeb, the Catalog of Fishes (via the previously existing Teleost Taxonomy Ontology), and the Paleobiology Database. The resulting curated taxonomy contains more than 106,000 terms, more than 104,000 additional synonyms, and extensive cross-referencing to these existing taxonomies. The Phenoscape Knowledgebase will leverage this taxonomic ontology by allowing for phenotype statistics to be displayed by taxon, including coarse measures of the extent of annotation coverage and phenotypic variation. Though phenotypes may be annotated to a species, the use of an ontological framework for the taxonomic hierarchy facilitates aggregating phenotypes to higher levels, such as genera or families. In the future, we hope to be able to integrate other excellent and rich sources of taxon-specific taxonomies, such as that in the Reptile Database or the International Ornithologists’ Union Bird List. This is a work-in-progress and the Phenoscape team is certainly interested to integrate new taxonomic sources as well as explore different ways that such a resource can be used and developed by the larger community.
In an effort to expand the user community and to demonstrate what is possible using our infrastructure, members of the Phenoscape team gave multiple presentations across two continents on our recent developments. In late October Paula Mabee gave an invited presentation on mapping phenotypes across phylogenies at the Muséum national d’Histoire naturelle in Paris. This was followed by presentations at the 73rd annual meeting of the Society of Vertebrate Paleontology (SVP) in Los Angeles and the 2013 meeting of the Taxonomic Database Working Group (TDWG) in Florence, Italy. Phenoscape had a significant presence at SVP with both a poster presented by Alex Dececchi demonstrating our progress in generating supermatrices from our annotations as well as a talk given by collaborator Karen Sears, using EQ supermatrices from Phenoscape fin/limb data to examine integration patterns across the fin to limb transition. Karen’s talk marks the first of the collaborations coming out of our 2013 San Francisco workshop. It also showed how data from Phenoscape can drive independent projects and is easily integrated with existing phylogenetic and statistical tool such as Mesquite and various R modules. The talks and poster were well received, with numerous researchers inquiring on how they could incorporate Phenoscape or use ontology based annotations.
There is a wealth of phenotypic information in the evolutionary literature that comes in the the form of semi-structured character state descriptions. To get that information into computable form is, right now, an awfully slow process. In Phenoscape I, we estimated that it took about five person-years in total to curate semantic phenotype annotations from 47 papers. If we are to get computable evolutionary phenotypes from a larger slice of the literature, we really need to figure out ways to speed this up.
One promising approach is to use text-mining. This could contribute in a few different ways. First, one could efficiently identify all the terms in the text that are not currently represented in ontologies and add them en masse, so that data curation does not have to stop and resume whenever such terms are encountered. Second, one could present a human curator with suggestions for what terms to use and what relations those terms have to one another, speeding the process of composing an annotation.
CharaParser, developed by Hong Cui at the University of Arizona, is an expert-based system that decomposes character descriptions into recognizable grammatical components, and it is now being used in several different biodiversity informatics projects. Baseline evaluation results from BioCreative III showed that a naive workflow combining CharaParser and Phenex, the software curators use to compose ontological annotations and relate them to character states, was capable of identifying candidate entity and quality phrases (it outperformed biocurators by 20% in recall on average) but had difficulty translating those into ontological annotations. This first iteration workflow also was not yet reducing curation time.
In March, a small contingent from NESCent (Jim Balhoff, Hilmar Lapp and Todd Vision) visited Hong Cui’s group in Tucson. We talked through improvements to CharaParser and the curation workflow, brainstormed plans for a more thorough set of evaluation tests, began refactoring of the code so that it can be more easily shared across projects, and gained a better understanding of what features make a character difficult to curate for humans vs. text-mining. We made substantial progress on all fronts, and are looking forward to seeing how much improvement in the accuracy and efficiency of curation will be achieved in the next round of testing.
We are also pleased to report that the CharaParser codebase will now be available from GitHub under an open source (MIT) license.
It’s easy to get caught up in the details when developing infrastructure. You know it will be useful – because the grant application said so! But there’s so much engineering to do. And no matter how thoughtful and deliberate a process you follow to anticipate the needs of your future users, once they have a complicated thing in their hands who knows how they will actually use it.
Enter the Phenoscape Knowledgebase. After a heroic data collection push this winter, our next release of the Knowledgebase will contain millions of evolutionary phenotypes from throughout the vertebrates, linked to genetic phenotypes from human, mouse, Xenopus, and zebrafish, and a particularly rich set of annotations for skeletal features of fins and limbs. The Knowledgebase is far from comprehensive, and annotations do not capture the full richness of the original characters in the evolutionary literature, but we think it’s a pretty useful resource.
So, it’s time to see what capabilities our users are excited by and what limitations frustrate them. To that end, we brought a small group of experts who look at phenotypes in a variety of different ways (e.g. genetics, systematics, evo-devo, clinical biomedicine, paleontology, even zooarchaeology) to the California Academy of Sciences in February, and we asked them what questions they’d most like to address using the KB as it exists today.
To help us in tapping into the assembled brainpower, we enlisted KnowInnovation, facilitation pioneers that specialize in helping researchers self-organize into teams to tackle creative research challenges. This they did with amazing resourcefulness, milking ideas out of us that we wouldn’t have imagined we even had. The workshop was no ordinary parade of PowerPoints. We did speed-dating to toss research ideas off of each other, generated a staggering number of post-it notes, sculpted creatures and skeletal parts out of clay and engaged in a host of other seemingly contrived but strangely liberating activities. We watched in amazement as Karl Gude took visual minutes.
And we came up with some great collaborative ideas for research that take leverage the Knowledgebase to ask questions that would have been difficult to impossible to answer without it, including questions about genetic convergence and parallelism, global comparisons of intra and interspecific phenotypic variation, and the evolution of phenotypes affected by duplicated genes. These projects will now serve as driving applications for Phenoscape so that we know better what our users really need the Knowledgebase to do for them. We look forward to reporting on the outcome of those in due course.
A big thank you to David Blackburn and the Cal Academy for providing such an inspiring venue, being exquisite hosts, and for conveniently having an open museum night during our workshop. Thanks also to a great group of participants and facilitators, and to to NSF for a supplemental award that helped to make the workshop a success.
At the end of October 2012, the working groups of the Phenotype Research Coordination Network (RCN) all met at the Asilomar Conference Center, in Pacific Grove, CA. One of the groups, the Vertebrate working group, made it their goal to discuss methods of representing phylogenetic and serial homology in anatomy ontologies, an issue that is central to Phenoscape as well. Though common ancestry is implicit in the semantics of many classes and subclass relationships (see for example the ‘homology_notes’ for digit in Uberon), most multispecies anatomy ontologies, including Uberon, VSAO, and TAO, do not assert homology relationships between anatomical entities. Nonetheless, homology is central to comparative biology, and therefore to enriching computations across data types, species, and evolutionary change.
The Phenoscape project had a strong presence at the largest Vertebrate Paleontology/Comparative Anatomy conference in the world this year, the Society of Vertebrate Paleontology annual meeting. In one of the large conference halls, and in front of a packed audience, I gave a talk on the history, goals and background of the Phenoscape project (“Phenoscape: A New Anatomical Ontology of Vertebrates”). The authorship also included Paul Sereno, Paula Mabee, Todd Vision and Hilmar Lapp. The talk was well received, and several attendees expressed great interest in our work. The difficult part now is to make sure this first spark of interest is maintained – this can be difficult when the community has not been exposed to ontologies before and the project appears to be so different from anything they have done before – but we’ll do our best to stay in contact with those people that expressed strong interest.
Alex Dececchi presented a poster on Phenoscape at the same conference (Phenoscape: bridging the gap between fossils and genes – his co-authors were J. Balhoff, W. Dahdul, N. Ibrahim, H. Lapp, P. Midford, P. Sereno, T. Vision, M. Westerfield, P Mabee and D. Blackburn), making sure that even those that could not attend the talk would get an opportunity to learn more about our exciting work.
Nizar Ibrahim, University of Chicago