Automated methods to score phenotypes in model organisms continue to develop and will permit previously inaccessible areas of biology to be probed.
Once you have the genome of your favorite creature sequenced, genome manipulation tools developed, and libraries for gain- or loss-of-function studies generated, then what? Especially if your inclinations run to genome-wide studies, you need a good, fast, reliable way to identify the alterations in phenotype that result from modulating or eliminating genes. This is also true for classical mutagenesis or forward-genetic studies: the likelihood of identifying interesting mutants depends on the scaleability of the phenotypic readout and the precision with which it reports on the process of interest.
Although many large-scale studies have been performed manually over the last decades, either through effective design of the phenotyping assay or thanks simply to the sheer doggedness of the scientists involved, there is little doubt that the development of automated or semiautomated phenotyping approaches could enable studies that have previously not been possible.
Indeed, several researchers are coopting technological developments in other fields and bringing them to bear on the phenotyping of model organisms. In the pages of Nature Methods alone, such strategies have ranged from the use of new (or newly applied) instruments that vastly increase the rate at which organisms can be evaluated to the application of computer vision–based tracking methods to monitor phenotypes that are too complex to be reasonably scored by human observers. Light-microscopic imaging of cell positions and lineages, and of fluorescent reporter gene expression—all within the context of the living organism—are also becoming amenable to automated approaches, which should permit analyses of growth and development that, without large-scale datasets, would be difficult to achieve.
It will be of interest to watch as these methods are put to work and refined in response to problems that may arise. And we predict that entirely new approaches are still in the making. Surely all the interesting phenotypes have not been studied yet.
About this article
Cite this article
de Souza, N. High-throughput phenotyping. Nat Methods 7, 36 (2010). https://doi.org/10.1038/nmeth.f.289
This article is cited by
Making deep neural networks right for the right scientific reasons by interacting with their explanations
Nature Machine Intelligence (2020)
BMC Genomics (2015)
Comparison of digital image analysis using elliptic Fourier descriptors and major dimensions to phenotype seed shape in hexaploid wheat (Triticum aestivum L.)