Technologies for quantifying plant development are underdeveloped relative to technologies for studying and altering genomes. As a result, information about plant gene function inherent in mutant phenotypes or natural genetic variation remains hidden. Even when not hidden, morphological data is often not in a form compatible with computational analysis.
Our Approach to a Solution
Processing electronic images with computer algorithms can render some important aspects of plant development into a numerical format compatible with bioinformatics analysis. For example, the midline of an object, such as a root or a stem, contains much information about the size and shape of the object. Quantifying midline length and curvature distribution is one means of quantifying the size and shape of the object. Currently, key aspects of root gravitropism and seedling photomorphogenesis are measured automatically or semi-automatically using this approach. To give genome-scale impact to the machine-vision approach, throughput of acquisition and analysis must be increased. Two approaches to this end are being pursued.
Robotics to parallelize the data acquisition, grid computing to distribute the computation
Robotically moving a camera between samples such that each is visited and imaged at the desired time interval would increase the rate of data acquisition many-fold. A motion-control gantry consisting of x, y, z computer-controlled linear slides capable of moving a pair of CCD cameras over a 1 meter x, y range with positioning resolution of approximately 10 microns is in a beta testing stage. You may watch a short movie showing the robot in action. The plant samples are held in an LED-illuminated 6x6 grid of Petri plates so that 36 experiments can be run in parallel. You may view a simplified diagram of the workflow. The Center for High Throughput Computing and the Condor software project are key partners.
Enabling the community to perform computer-vision-based experiments on plant development
Increased throughput will also be achieved by enabling many members of the community to perform their own morphometric experiments. We have prepared a detailed guide to setting up an image acquisition apparatus. The Phytomorph project will also provide access to some image-analysis functions that operate on time series of digital images. In some cases, uers will upload images and receive results electronically. The HYPOTrace analysis program may be downloaded as a stand-alone executable.