Massive database of 182,000 leaves is helping predict plants' family trees
The technique could be used on everything from flowers to cells to examine the factors that influence the shapes of plant parts.
The story of a plant is etched in its leaves. A tree growing in a cold environment with plenty of water is more likely to have large leaves with many serrated teeth around the edges. But if the same species lives in a warm, dry region, its leaves are likely to be smaller and smoother.
Now, an atlas that traces the shapes of 182,000 leaves from 141 plant families and 75 locations around the world shows promise for refining scientists’ ability to read that story. Using that atlas, researchers found that leaf shape alone accurately predicted where a leaf was collected 14.5% of the time, and plant family correctly 27.3% of the time. That is far better than predictions made using conventional methods to describe a leaf's shape.
Researchers hope that the approach will help them to learn more about the forces that shape plant leaves, and even to get a glimpse of ancient climates by analysing the shapes of fossilized plants. “It’s an amazing data set,” says Dan Peppe, a palaeobotanist at Baylor University in Waco, Texas. “We’re getting closer and closer to automating measures of leaf shape, and using that to figure out the taxonomy of a plant and reconstruct climate.”
The results were posted on 20 June to bioRxiv, a server that hosts biology preprints. Plant morphologist and lead author Dan Chitwood also presented the study at the Botany 2017 meeting in Fort Worth, Texas, on 27 June.
Shaping up the data
Chitwood, formerly of the Donald Danforth Plant Science Center in St Louis, Missouri, and his colleagues pulled together data from their own work on specific plant groups, such as grape and tomato plants, as well as several large data sets from projects that aimed to catalogue a wider swath of plant species and locations.
They then used a topological method called persistent homology to analyse the shape of each leaf. The method assigns each pixel in an image a value according to the density of the pixels around it. The team broke each leaf into 16 parts, and analysed the pattern of values in each one. The researchers used the resulting catalogue of leaf shapes to look for taxonomic and geographic relationships between species.
Chitwood’s ultimate goal is to reconstruct the leaf ‘morphospace’: the full catalogue of possible leaf shapes. “If you could measure all the leaves that currently existed and all the leaves that ever existed, would it be completely random?” he asks. “Or would there be some leaves that never showed up? Was it because the plants can’t make them?”
Persistent homology has been used to map everything from networks of neurons to the structure of musical phrases, and Chitwood hopes that it can provide a unified method for analysing all parts of a plant. Others are eager to apply the same method to their own research needs. Plant morphologist Yannick Städler of the University of Vienna wants to use the technique to analyse his growing library of X-ray images of flowers. He hopes that it will help him to overcome a stumbling block with conventional morphological methods, many of which involve placing landmarks — points on structures that recur across species — on images.
Those techniques work well for animals, he says, which tend to have obvious landmarks: the point at which two bones meet, the corner of an eye, the tip of a nose. But flowers often have smooth, curved surfaces, which makes it difficult for researchers to pinpoint specific landmarks. “This has been a horrible problem in leaves and in flowers,” Städler says. “It has held us back.”
Leaf by leaf
Palaeobotanists such as Peppe are hoping for ways to automate the analysis of fossilized leaves — a process that currently requires painstaking work to manually place landmarks on fossils for analysis.
Other projects are analysing plant features, including leaves, fruits and flowers, to enable researchers and hobbyists to rapidly identify them in the field. A project called Pl@ntNet, for example, has collected millions of images submitted by users around the world through a mobile-phone app, says botanist Pierre Bonnet of the French Agricultural Research Centre for International Development in Montpellier, France. So far, the project has analysed 580,000 images from 13,000 plant species using machine-learning techniques.
Pl@ntNet is better at identifying plants than Chitwood’s atlas, says one of Bonnet's collaborators, computer scientist Alexis Joly of the French Institute for Research in Computer Science and Automation in Montpellier. The team hasn’t yet used Pl@ntNet to study the diversity of leaf shape, he adds.
Chitwood hopes to feed the results of his topological methods into machine-learning algorithms as well, to see whether those can improve his taxonomic and geographical predictions. But he is more interested in understanding leaf shape than in classifying plants, he says.
There was a time when it seemed as if such efforts to understand plant morphology were dying out, says Städler. But the field is experiencing a renaissance thanks in part to widespread efforts to characterize the traits of plants — particularly crops — and to understand how genetics and the environment influence them.
“Morphology is being reborn,” Städler says. “That’s where the field is headed. And I think, especially together with genetic data, we have a very bright future.”
News
Need Assistance?
If you need help or have a question please use the links below to help resolve your problem.