If scientists better understood how the many thousands of genes in a plant affect growth, development, and physiology, it would be easier to improve crops. Scientists would also better understand how non-crop plants fundamentally support life on this planet. It is relatively straightforward to characterize a plant’s genes (the genotype) these days but the growth, development, and physiology (comprising the phenotype) are much harder to measure. Yet, understanding how the genotype produces the phenotype requires good measurements of both. Digital imaging has emerged as a methodology for efficiently measuring many important phenotypes. This project will address three general deficiencies that prevent more phenotypes from being measured efficiently by image analysis. Pipelines: not enough computational pipelines capable of measuring broadly useful phenotypes have been constructed. People: not enough people interested in plant biology research are trained in the computer science and data science methods to make more pipelines. Process: that art and science of combining pieces of technology to form a pipeline that a biologist can use as a tool is an interdisciplinary process that is currently difficult to replicate everywhere it is needed. If 20 years ago the public generally understood how computers and imaging could be applied in plant biology research, this project may not have been necessary. Therefore, a built-in robust outreach component will present concepts and engaging exercises to youth outside of a university setting. Collectively, the project will advance the field of measuring plant phenotypes.
Many phenotypes including those related to crop health and productivity can be measured from images using computers and applied math techniques. This approach to measuring phenotypes is presently limited by too few automated and high-throughput image analysis pipelines, too few people in the community capable of creating them, and no systematic process to facilitate new pipeline creators. This project addresses these deficiencies by creating effective high-throughput image analysis pipelines, training postdocs and students in the necessary technical domains and soft skills, and by making the pipeline creation process as formulaic and modular as possible. With input from community surveys and an Advisory Board, a prioritized list of needed measurements will be created. Pipelines designed to meet those measurement needs will be deployed as Web services on public cyberinfrastructure such as CyVerse and the Open Science Grid. Postdocs and graduate students will be trained in image analysis techniques, machine learning, virtual machines and containers, file transfer and storage methods, and high-throughput computing. They will learn to abstract biological imaging problems and adopt the perspective of the user to produce new robust measurement tools. In cases where a good solution exists but is not readily accessible, the trainees will deploy the existing tool on public cyber infrastructure. The overall result will be more high-throughput image analysis tools to meet the community’s phenotype measurement needs and an increase in the community’s capacity to create more. This award was co-funded by the Plant Genome Research Program in the Division of Integrative Organismal Systems, and the Infrastructure Capacity for Biology Program in the Division of Biological Infrastructure.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
