Formalizing how plants live and adapt through algorithms

What is your favorite word?
It's Woensdag - it's Dutch. And when I learned Dutch, I always wondered, what are the ‘Woens’? I saw them as little animals or something like that in my mind. When I asked in my Dutch course, they told me it's coming from the day of Marcus.

There's a whole story about it, but I always thought about little animals that are ‘Woens.’

What's your favorite breakfast food?

Pizza from the day before.

Money or happiness?
Happiness, definitely, no question.

Your research is motivated by the impositions that climate change creates on the agricultural and natural plant ecosystem. Can you tell us how you became interested in this field of study? And can you also tell us what a phenomicist is?
Originally, I studied computer science, but I always liked nature. I tried to combine it all the time. I had concerns as a person outside of academia, about climate and nature, how it changes over time now, combined with my interest. That merged into what we now call a phenomicist. It’s a new word we just invented in a working group in the International Plant Phenotyping Network.

We try to define what we are when we do things outside, when we start measuring the plant.

How do we define ourselves? That was the question behind it, and we think now in terms of like, ‘Okay, a phenomicist is making formal systems to study plant life,’ right? And to describe how plants interact with the environment, and how they change suddenly? Do they survive? So, we are making formal tools to study that, and I think that's how this word exists. We started using it about half a year ago within this working group and putting it into grants.

A physicist, you know what they're doing, they're doing the physics of the planet or the universe. We are now doing the phenomena of plants as phenomecists by using formal systems. Informal systems are a very broad term in the way we define them. It could be an imaging system, a hardware that is defined based on whatever electrical circuits, right? This is a formal system, but it can be an algorithm. It can be a simulation. It can be an imaging system that you just combine an algorithm halfway, or electronics could be everything; it's very broadly defined.

I know genomics, but what is phenomics?
The first thing to recognize is the appendix, which is omics. It's talking about tools, and phenomics deals with everything that is the phenotype. For me, that's plants.

What does a plant look like? And how do I quantify what it looks like? That could be like the length of a branch, the number of leaves, or the area of leaves. And then, in my case, I go below ground: how many roots, or single roots, are actually in these root systems? Roots are also just like a lot of branches, like taking a three-crown upside down. You can measure length, diameters, angles, but you can also, in our case, try to find ways to summarize all these diameters and angles into one mathematical expression. It’s a way you can make a whole system of branches comparable to each other. If you have either single measurements like angles or these summary descriptions, you can start, for example, using genomics and link it to genes. 
 

You can say, ‘Is this angle, this shape of this whole plan? And is there any genetic control?’ Then you say, ‘There's a genetic control to it. This shape also seems to be very frequent in a desert landscape, and you have a relationship to the environment.’ You can make this link. How do these plants survive in these very extreme environments, or even normal environments? It's just a relationship that you can make. Then, from the phenotype, what you see is how plants live in an environment.

Where did you study to obtain your degrees?
I got my bachelor-master’s degree in Cottbus, Germany. It’s southeast of Berlin, towards the Polish border.
 

Then I went to Delft, the Netherlands, which is on the other side of Europe. It's close to the ocean, and that's where I did my PhD on trees and how tree crowns are structured, and made the first algorithm to analyze complete tree crowns.
 

Back then, there was no field like we talked about now, like phenotyping, there was no phenomicist. There were a few people who started looking at plants and how to measure them in general.

We started doing a little algebra and found algorithms that take every branch out of a tree and make it measurable. That was the challenge back then. In my master's degree, we started making the first surface reconstruction of a barley plant. You see, in the 3D models in television, in animation videos, if you look at the plant, it becomes inherently more difficult. It's 3D, and then you have leaves, and they're flat, they're kind of 2D in your data. If you turn it into data, it becomes two-dimensional. If you look at a little grass, it becomes like a line in your data. There will suddenly be just a one-dimensional line. So we have challenges in dimensionality.

If you go outside into the real world, it becomes a whole different story. Then you have your instrument that you use to measure a plant interacting with its environment. You shoot a laser, which is light, and put it on a water drop. It has a size, and it hits just one part of the plant, a piece of the leaf, and then the other piece hits the stem of a tree. You get some measurement in between. It becomes very noisy and messy compared to what you see in the movie theaters.

Then you see your 3D animation movies. That's all, it will be put in either 3D imaging, and then we start measuring these plants. But with all of that noise and the messy chaos around makes sense out of all of that. What do you do with all of these measurements? If you have a lot of plants measured, we can now relate that, for example, to genes. We can say: look at the length of this branch or the size of the root that we get, always related to a certain gene, or several genes that the plant has. Then you link this trait that you measure, which makes more yield under drought, like here in Tucson. If we know these genes, we can tell breeders who work with them to get more yield somewhere in the middle of the desert. That's where it starts getting very interesting, but it's the step before making the plant right that's important to know.

So what was it that brought you to the University of Arizona and the BIO5 Institute?
That’s another long story! Back in my postdoc at Georgia Tech, I collected a lot of the data I work with now in Wilcox, Arizona. In Wilcox, Arizona, there was a Root Research Center for a long time. That's where I collected my data initially. I got a job in Georgia, at the University of Georgia, UGA, back then, as there was no position here in Arizona.

I collaborated for a long, long time with the folks of CyVerse here at BIO5. I started my first professor position at UGA. I am still working on that data and still collecting data here. Then there was a position free here.

Now I’m closer to my data. I don't have to fly out, I just drive 20 minutes from my house to the field. All my lab is sampling in the field in Wilcox right now.

Have you had any cool breakthroughs lately in the lab that you would be willing to share with us?
So it takes a while to build up and seems like a breakthrough to the public, but it's been building up for a long time. Over the last 10 years, what has became very clear to us, is that if we look at how plant breeding is approached, it was always about this one trait, this one feature that we want to improve, and that if we go out in the field, we just take a few plants and we assume everything is the same, right? We look at it, we measure it, and all these measurements are more or the same on all plants that we look at.

But then we said, ‘Okay, let's look at a whole field.’ We have all these nice methods, these formal systems that we build. We can look at so many images, we can simulate things that we cannot see. Just look at a whole field, what's going on there below ground, and we see that there's a lot more variation. There's so much stuff out there. There are small roots, big roots. There are roots with very flat angles, deep angles, all in one population, so in one field, that should be the same genes with the same variety all the time. You don't expect so many changes between them.

Then we looked at whether we can group that, or put that in different groups. Could we compute something out of this data and link it to certain functions? It turned out they're all specialists. We think they help each other by being specialists in these extreme climates, like here in Tucson, it's a desert. You want to be good at something and then help others. I think that's kind of the bigger thing we found out right now.

We have another exciting finding that is more of a side project of four, five years. The main focus initially is that we've also found a kind of a little hair. Roots have hairs that help with the uptake of nutrients from the soil, and they're very much extending the surface that is in contact with this. They were always thought to be at the end of each root, called a root branch, but we found them very early on. Now in development, they were at the end of these long roots that are a bit older, and now we found them in the first few days, but they look different. They just had a very different shape. So we put it under the microscope, and we looked at the shape. It's different from root hair, but it's also different from any other little hair we find on the ground.

So the question is, what is that? It seems also that it has to do with nutrient uptake in the very early days when the seedling is established. And then we looked around a little bit, not fully, but it doesn't seem to be in every plant. It seems to be very specific to a common being. These beans have that, which makes sense. Now, they're also from around here, so if you think about Wilcox, there's the Bonita Bean Company. Beans are very popular here. When the seedling establishes, it's very vulnerable, basically like a baby. And if you put it in an extreme environment, it easily dies. We sometimes have up to 20% seedling death. If you have this in some plants, but beans somehow have this hair that helps them, just one cell that extends out very early on, like in the first five, six days already. We did the first genetic studies on that, where we saw it's somehow related to either nutrient uptake or a little bit to defense against pests. There's also some antimicrobial function involved.

What are the implications of these discoveries for society?
It's about plants and how they survive in extreme environments, and that's the climate change link.

Plants, with all the services that they provide for us, if you look at crops, it's food, but it can go further. You can build houses out of plants, or we did that a long time ago. Still, many houses in the U.S. are built with a lot of wood. It's also an energy source, right? You can burn wood or make biofuels out of it, maybe medicinal plants that you want to use. It's generally, how do we get any kind of product from the plant in extreme environments?

Can you share how the computational imaging and simulation techniques you develop are applied in the lab and the field?
We make images. One thing we do, if you go into the field, is we have scanners. We put a plant inside the scanner, it takes a lot of photos from all sides of the plant. Then we take all these photos, put them into an algorithm, and the algorithm gives us a 3D model back, and that's the basis on which we measure. There's a difference in how the imaging instrument looks. It's not as robust in the lab. It can be any kind of photo camera, you know, whatever you have at home, it could be iPhones. How it looks in the field is that we have to dig up these plants, and it gets messy. The problem that we face with imaging, especially in the field, is that the soil has a lot of iron, and if you think about either electromagnetic waves, like whatever X-ray or any other wave you want to send down there, your resolution gets very low. We are still digging them up in the lab.

You have other possibilities. If the volume is very small, you can put, in a controlled way, a lot of energy there. You can put it from the side, right? There's the soil, and your plant might be in the soil. It's hard to get from the side without destroying too much.

I think they are the limitations that we just face in the field. We are starting to dig them up, wash them with water. We say, “hey, just take a lot of plants, like 1000s,” and then we destroy a lot, but we have a lot of them. So we can start making statistics because of the number of plants. We have to just look at very many, and then we figure out what is valuable in this data and what is not. Then you have just like three, four, or five samples, we have 1000s. That's how we attack the problem. We just look, what do we see over and over and over again, and it’s also an incomplete field, right? This is just not small. You'd need 15 to 20 people to process 5000 plants in two weeks. Think of that, that's hard work. You're going out with a shovel, you're digging them up to get them out of the ground.

You have to be very careful not to sever the root!
It is kind of a standard way to destroy the root like that. We are not claiming that we get the whole root out; that is impossible. But you can say you're getting a good part of it out of the ground. We can measure certain things consistently, like we can measure an angle, we can measure a diameter at this root. We certainly cannot measure the length reliably or the depth, because it's limited by how deep our shovel goes. That's just something to take into account. What is the data telling you?

I think this is often also a misconception out there, that if you are destructive, you cannot measure anything. Especially if you talk to researchers in the lab who are just trained in lab environments, they have these perfect systems in the field. It's not perfect, right? Their plants are dying in between, you're living with this incomplete data, but we try to get information out of incomplete data.

Digital imaging of root traits or DIRT. Is that one of the techniques that you use when you're measuring? This is the platform that your lab created, and I think it was at your lab in Georgia that created this.
I created that during my postdoc over 10 years ago.

We still use DIRT to some extent. For example, if you look at the specialization of fruits in a whole population, a whole field, as I described before, the first results we have are still made with DIRT, and it's based on how you can summarize a root system in one value, one construct. It's a bit of an abstract thing to think about, but we look at the shape of the root, and we try to find only one thing to describe it. And then we can take this one thing very, very often and start finding groups.

For example, it's like you find more round shapes, more triangular shapes, more square shapes, or oval, and that's where we are with these whole field studies right now. But if you look a little bit into the future, I also mentioned a scanner. That's where we are right now. I mean, we start bringing that to these massive throughputs in the field. It takes a lot of technology development to get the throughput, simply for 1000s of plants in two weeks, and we are doing the first study in 3D on that level, but we don't know what comes out of that.

DIRT is still there, but it's also 10 years old. So we are now already having DIRT 3D, it's already published as a system, just the scale to the field. This is now the first time that we have gone to 5000 plants.

We already have little studies of like, four or 500 plants on sorghum that we did last year up in Maricopa Agricultural Center, and we have also maize or corn studies out there to just distinguish genotypes from each other, like classic phenotyping. We just look at many different varieties, and then we just look at the differences between the varieties, where you have two, three samples per variety that are out there. It's not looking at the whole variation. It's a pre-sample, where you go out and say, ‘I'll take five plants and then the three most similar ones I accept as being the phenotype.’ That's kind of the classic way to do it.

Do you have any collaborators on campus or elsewhere who have contributed to the success of your research, or has your research contributed to the success of any collaborators?

So many here on campus. I think the biggest study I did until now was with Giovanni Melandri here at BIO5. That was the sorghum up at Maricopa.

For many years, I have collaborated with Jonathan Lynch and Penn State. I think we helped each other evolve our research. He's the one who brought me into root research at the beginning. When I came here, I was in a project where he was involved in, and that's where I met him. It was my postdoc project, and since then, almost 15 years. He's retiring now, maybe you never know. He tried to retire a few times, but I think he also likes to keep on working. He's very, very passionate, and he's an amazing researcher out there. But he’s also a mentor in some sense, because he brought me during my postdoc to South Africa, and then we did studies there. I just started learning about roots and why roots are important, and how they can feed the world. He was a big, big influence in getting into the root phenotyping and applying technology to roots. He started digging up roots, but all the parts that he could scale up and do the imaging were in collaboration for the first time with him for me. There are many more collaborators. I have a collaborator in Thailand that I met in South Africa on that farm very back when we started doing whatever student exchange with each other. That's Quan in Thailand. There are many more collaborators over the years that I have had.

Are the collaborators interdisciplinary kinds of labs?
When a student comes into my lab and I explain what our lab is, I tell them our lab is not the lab you normally see.

It's not like you're coming into a molecular biology lab, or everybody knows PCR or any kind of standard technique. We have a mathematician, computer scientist, cell biologist, and molecular biologist. And even then, if you go down to the undergrad level, even the undergrads are unique. We have a biosystems engineer and physiologist. Everybody's the boss in their own field, and they often know much more than I do. That's different from many of the typical labs where one person could take over for the other. That's hardly possible. Of course, we have a overlap, but not too much, and the lab is interdisciplinary.

My background is not even in biology, but that’s what happens then when you collaborate. We often collaborate with very specialized labs that do one thing. We help them with the breadth of different disciplines, and we get a lot of expertise, very specialized expertise, that they often cannot cover ourselves.

Do you have a mentor or a few mentors who have impacted your life?
There's certainly a whole list of mentors from all kinds of fields. The typical ones are your PhD advisor and postdoc advisor. But the reality is, everybody is your mentor; your whole lab, everybody, if you're willing to listen, everybody can be your mentor.

You always learn something from the undergrad to the big professor that you meet at a conference. You just need to be willing to listen. So I think everyone is a mentor, but some people hire you as a postdoc or as a PhD who influenced you more than others, because it's just the time they spent with you.

What is your why? Why do you get up in the morning? What keeps you going?
The nice thing about our job is its dichotomy.

You come home as a scientist, and you have a new problem every day. Most people don't come home with a problem every day. For us, that's the thing, but it's also the progression every day that you can make a step towards a little bit of a better world. Either it's the food problem, how do we feed people under this climate pressure? Or get more energy under this climate pressure, producing plants? There’s always a little step that you make towards a better world that is nice, despite having a new problem that you bring home every night.