Meeting Abstract
Recent advances in phylogenomics and next-generation sequencing technologies made phylogenetic inference of large radiations of organisms possible. These large phylogenies have been successfully used in conjunction with existing comprehensive datasets to answer key questions about species diversification and morphological evolution. However, collecting large amounts of new phenotypic data has typically been bottlenecked by researcher availability and effort. For geometric morphometrics in particular, a single individual often collects shape data to reduce methodological measurement errors. Here we present a method and toolkit to efficiently collect two-dimensional geometric morphometric phenotypic data at a “phenomic” scale using workers recruited through Amazon Mechanical Turk. We examine inter-and intra-observer accuracy by assigning identical image sets and digitization protocols to experienced fish morphologists, undergraduates, and Amazon workers, and compare these data to a “gold standard” set of digitizations. Our results show that the quality of Amazon workers’ data are not significantly different from results collected via traditional sources and thus are a viable resource to quickly and accurately collect large amounts of phenotypic data. We also have developed a pipeline that streams crowdsourced data from the web and can iteratively analyze and update results as new data arrive. We demonstrate this workflow by examining body shape evolution of 539 species in 7 families of ray-finned fishes (Acanthurid, Apogonidae, Balistoidae, Chaetodontidae, Labridae, Pomacentridae, and Tetraodontidae) and discuss the relationship between their diversity and phenotypic disparity.
