Author
|
Submitted to: Rice Technical Working Group Meeting Proceedings
Publication Type: Proceedings Publication Acceptance Date: 12/22/2015 Publication Date: 7/2/2017 Citation: Edwards, J. 2017. Free and open source tools for high throughput rice phenotyping and data storage. Rice Technical Working Group Meeting Proceedings. March 1-4, 2016, Galveston, Texas. CDROM. Interpretive Summary: Technical Abstract: Commercial automated phenotyping systems are cost prohibitive for many research programs. Free tools can be repurposed to provide automation when integrated into a conventional plant phenotyping program. These include barcoding for identification of individuals, semi-automated image processing to evaluate plant growth, and streamlined field data collection to populate phenotype databases. We have evaluated several open source tools in plant genetics research activities, and will discuss the implementation and results. Barcoding is useful for tracking materials, preventing misidentification, and rapid data collection. Items that may be labeled with barcodes include seed envelopes, pots in the greenhouse, field plots, petri plates for germination, tissue collection containers, DNA samples, PCR primers and PCR plates. Barcodes may be the conventional 1D type or 2D QR code type, and each has advantages and disadvantages. The 1D barcodes can be read with standard laser-based barcode scanners. The length of the 1D barcode increases linearly with the amount of information stored and may not fit on seed envelopes or tubes. 1D barcodes that become damaged cannot be read. 2D barcodes can store substantially more information for the amount of space they occupy. They have built in error correction so that even if part of the barcode is damaged, it may be still read. The level of error correction in a 2D QR code is adjustable, and use of the highest level or error correction is recommended for non-optimum conditions such as greenhouse or field use. 2D barcodes require a device with a digital camera to read such as a smart phone or tablet. We have successfully used barcoding to assist in rapid inventory, verifying the correspondence of seed pack and pot, or plant and collected sample, and image identification. Image processing techniques can be used on large numbers of digital photographs taken of plant growth over time. Processing these images and extracting useful information is challenging, and automated methods are needed with minimal levels of user intervention. One application of image processing is to estimate the leaf area of a plant as a non-destructive proxy for above-ground biomass. The GNU Image Manipulation Program (GIMP) is useful for point and click operations such as selecting only the plant leaves in an image and removing any non-leaf background. Pipelines using the Linux command line tool ImageMagick have been developed to further process many images simultaneously to calculate pixel area of selected regions of the image. A 2D barcode contained in an image can be extracted using the ZBar barcode reader and corresponding package in the Python language. This is useful for automatically identifying and renaming images based on the contained barcode information. The barcode can also be used for calculating the area of a known object in the image (e.g. the 2D barcode itself printed to a standard size) for use in converting pixel counts to area-based measurements in known units. Data collection and storage is another challenge that can be addressed with open source tools. We have implemented a database for rice breeding and genetic research (Ricebase). This is a relational database that inherits its architecture from the publically available code of the Sol Genomics Network (SGN) database. The database schema includes the community developed Chado natural diversity module. Breeders tools include pedigree tracking/parsing/drawing, storage of accession related phenotype data and molecular data including conventional molecular markers and large-scale next generation SNP data. A genomic selection module provides breeding value predictions directly from the database when phenotypic and molecular data are available. Experimental designs are integrated with the Android tablet app Fieldbook to directly import and export data collected o |
