Researchers at the John Innes Centre for plant and microbial science were looking for a cost‐effective phenotyping platform for automated seed imaging. They figured a machine learning-driven image analysis was the quickest way to deliver this essential, yet challenging, aspect of agricultural research. Sounds complicated, but they found that our tiny computers could handle it all.
What is phenotyping?
A phenotype is an organism’s observable characteristics, like growing towards the light, or having a stripy tail, or being one of those people who can make their tongue roll up. An organism’s phenotype is the result of the genetic characteristics it has – its genotype – and the environment in which it lives. For example, a plant’s genotype might mean it can grow quickly and become tall, but if its environment lacks water, it’s likely to have a slow-growing and short phenotype.
Phenotyping means finding out and recording particular aspects of an organism’s phenotype: for example, how fast seeds germinate, or how broad a plant’s leaves are.
Why do seeds need phenotyping?
Phenotyping allows us to guess at a seed’s genotype, based on things we can observe about the seed’s phenotype, such as its size and shape.
We can study which seed phenotypes appear to be linked to desirable crop phenotypes, such as a high germination rate, or the ability to survive in dry conditions; in other words, we can make predictions about which seeds are likely to grow into good crops. And if we have controlled the environment in which we’re doing this research, we can be reasonably confident that these “good” seed phenotypes are mostly due not to variation in environmental conditions, but to properties of the seeds themselves: their genotype.
Growers need seeds that germinate effectively and uniformly to maximise crop productivity, so seed suppliers are interested in making sure their samples meet a certain germination rate.
The phenotypic traits that are used to work out whether seeds are likely to be good for growers are listed in the full research paper. But in general, researchers are looking for things like width, length, roundness, and contour lines in seeds.
How does Raspberry Pi help?
Gathering observations for phenotyping is a difficult and time-consuming process, and in order to capture high‐quality seed imaging continuously, the team needed to design two types of hardware apparatus. Raspberry Pi computers (Raspberry Pi 2 Model B or Raspberry Pi 3 Model B+) power both SeedGerm hardware designs, with a Raspberry Pi camera also providing image data in the lower-cost design.
The brilliant team behind this project recognised the limitations of current seed imaging approaches, and looked to explore how automating the analysis of seed germination could scale up their work in an affordable way. The SeedGerm system benefits from the cost-effectiveness of Raspberry Pi hardware and the open source software the team chose, and that makes us super happy.
Read the whole research paper, published in New Phytologist, here.
Raspberry Pi in biological sciences
Dr Jolle Jolles, a behavioural ecologist at the Center for Ecological Research and Forestry Applications (CREAF) near Barcelona, Spain, and a passionate Raspberry Pi user, has recently published a detailed review of the uptake of Raspberry Pi in biological sciences. He found that well over a hundred published studies have made use of Raspberry Pi hardware in some way.