December 03, 2020

Putting tumors on the map

By Conor McKechnie and Dodi Axelson

Putting tumors on the map

Why can’t we predict who’s going to benefit from a cancer treatment, and who isn’t? Turns out we’ve been missing an important piece of the puzzle. Grab your explorer hat and pop it on your head, because we’re going mapping.


DODI: Conor, do you remember, there was an episode about what we have learned in biotech from other industries? And I chatted to Paul Goodwin for a quick second.

CONOR: Yes, I do. Hang on, refresh my memory.

PAUL: There’s a dye that's in your blue jeans, if you wear blue jeans. That dye is identical or extremely similar to the dye that pathologists use to stain the nucleus of cells. And there are pink dyes that are used for staining, particularly silk. It's very close related to how pathologists study the spaces between the cells in a pathology sample. A lot of these tools are borrowed from the textile industry.

CONOR: Okay, yeah. I remember now. You teased us that he would be coming back in a later episode.

DODI: Today's the day.

CONOR: Oh, brilliant. So, where's Paul taking us today?

DODI: Well, grab your explorer hat and pop that on your head, because we're gonna go mapping.

CONOR: I love mapping. But what are we mapping?

DODI: We're gonna start with the complexity of tumors. Can you imagine mapping a tumor?

CONOR: And so that's what matters on today's episode?

DODI: Right-o, off we go.

PAUL: My own training is as a physiologist.

DODI: This is our good friend, Paul Goodwin, science director for Cytiva.

PAUL: I live in Washington state in the United States.

DODI: And Paul is trying to map tissues and tumors.

PAUL: If I run out of gas on the road, and I need to find out what the closest gas station is, I pull up my cell phone and I say, “Gas stations near me.” And the report back is that there are 5620 gas stations in the state of Washington. It wouldn't help me know which way I should walk down the road to get gas. What makes it useful is it says the closest gas station is west of you by 0.6 miles, and they take your credit card. And while you're there, you can get Snickers™.

That’s the kind of information we don't have when we want to study tissues and tumors. So, I can get the information, how many Snickers they have, how many gas stations are there. But I have no sense of what that place is. Unless I look at a map, where I find out where the roads are. But I don't know where I am. I don't know where the gas stations are. I don't know where the Snickers are. And so putting that information together in a useful way allows us to get much more information about where we sit.

CONOR: I guess what he's saying is that understanding where tissues and tumors are will help us not just understand them, but understand what to do about them.

PAUL: Right. So, if I want to understand a tissue or a tumor, I need to understand what are the elements that are there, what are the pieces that are involved. It's not enough to deconstruct it into just a bunch of pieces. It is also the location of those pieces relative to each other that makes a tumor difficult to treat. And so when we've taken tumors apart and grown pure populations of cells in dishes, and develop chemotherapies against that, and then we go try to apply that back to a patient, they most often fail. Because we've failed to take into account the complexity of the tumor, which is just not a collection of those pieces. It's this organizational structure as well.

CONOR: Location, location, location.

DODI: As any good real estate agent will tell you.

CONOR: So, how new is this practice of mapping tumors and tissues? And what's the history behind it?

PAUL: Historically, pathologists have been around for 150 to 200 years, and they've looked at tissue and described it qualitatively. And then there has been a movement over the past 10 to 15 years towards very quantitative analysis of the pieces of the tumor. And the realization is, and this really has come about largely because of immunotherapies, in some patients immunotherapy is extremely effective. But that's only about 30% of patients when dealing with solid tumors.

And when we have a solid tumor and 70% of immunotherapies fail, well, why do they fail? And we realized that it wasn't just because of the cells that are in the tumor, it has to do with their microenvironment: How do they get blood? How do they get rid of waste? What cells are next to each other? What signals are they picking up from their environment? And those are all questions that involve place. Therefore, I need to map them to put into my analytic not just what's there, but where are they, and where are they relative to important structure.

DODI: So, in other words, it's pretty recent. We're still learning a whole lot.

CONOR: Okay, that's really interesting. We’ve talked about the Human Genome Project. And we’ve done a Discovery Matters episode on the protein atlas with Mathias Uhlén.

DODI: That's right, exactly. And so we've been all over the map of life science podcasting. Virtually speaking.

CONOR: So, is mapping tumors like just another stamp in our passport? Is it similar to these practices that we've talked about before?

DODI: Yeah, let's call them Twin Cities.

PAUL: The realization I had was that we had moved so far down the road towards understanding the genetics of cancer, understanding the proteins of cancer, understanding the signaling pathways, that we lost track of the fact that where they are in their environment is an essential component. And what I want to be able to do is search for place in my analytic. Just as I do today, I search for a gene or look for a protein’s expression, I want to include that structure in my analytic inference engine that tells me whether this patient is going to respond or not respond to a given therapy.

CONOR: So, it sounds like Paul's got a pretty high demand for what he wants to get out of this process. But does the overall industry agree with him?

DODI: Well, Paul says the industry is, in fact, turning.

PAUL: When I first started thinking of this topic a year ago, I think it was pretty wild. But now I see more and more meetings, seminars, papers being written on the need for what they call spatial analytics.

DODI: And the difficulty is that much of the work is done by hand.

PAUL: Some undergraduate or graduate student or postdoc or clinician goes in and draws structures or measures the distance from cells to blood vessels. There is no adequate descriptor that's been used to generally define what the map is of the tissue. And so what is new, and what I and others are starting to think about, is how do we draw on tools from our friends in geography, who have described place for hundreds and hundreds of years? And what other tools can we borrow on how they describe where the gas stations are along the highway, the environment where the trees are relative to a stream, use those sorts of analytical tools to combine along with the other -omics — the genomics, the proteomics — to more thoroughly describe the tumor, so we can understand if that tells us response to therapy, and who's going to respond and who's not going to respond?

CONOR: Well, okay, that is fascinating, isn't it?

DODI: It's like, shouldn't we be calling a cartographer? Well, we should be calling it a glacier. Because one day, you take a measurement, and the next day, it's not the same at all.

CONOR: Constant change is the only constant. So, what kind of biotech breakthrough are we looking forward to here? What does Paul expect that we can achieve once we can really accurately map tumors and tissue in this way?

PAUL: There are a couple of areas where I think this is going to be really important. I've talked a lot about oncology, about cancer therapy. One of the puzzlements that oncologist face is that when we apply new tools for treating cancers to cells where place or rigid location is not as important — to blood cancers, things like lymphoma and leukemia — that cell therapy worked remarkably well. The logical conclusion was great.

Now let's go take on the solid tumors because more people die of solid tumors than blood cancers. And those are made up of tissues. So, let's just take the same tools and apply it to that cancer. And we had this poor response: about 30% responded wonderfully, and 70% had no response at all. So, how do we take this information, and we're desperate for a way to be able to understand who's going to respond and who's not going to respond? We're not good at making those estimates today, because we don't have all the information in one place that describes both the location of things as well as what is there to be located.

CONOR: So, he's not like just a tissue and tumor mapper, but a tissue and tumor hunter.

DODI: True. Okay. So, let's open his hunting toolbox and look inside. What has he got?

PAUL: One of the things that's happened in the past 10 years is a realization that the average performance itself doesn't tell you as much as individual cells tell you. So, there's been a huge amount of effort the past five years to rather than describing the genes that are being expressed in the liver, to ask what is the variety of gene expression that occurs In each individual cell that might be in the liver? And that's called single-cell sequencing, or single-cell transcriptomics, where it gives you the ability to say, what are the cells? And what are they trying to do? And that's really, really important.

But if in the process of doing that, I lose location, then I've thrown away a vital piece of information to understand another piece of information. How do I bring that kind of analytics, understanding of where the proteins are, where the cells are, what cells are present, what structures are present into the context of the tissue itself? But we really believe that if we had a better understanding of that information, we'd have a better sense of what treatments are being effective and which ones are not, and how to predict efficacy before treatment. And so the tools that are being used are single-cell sequencing, looking at protein expression.

And a really big topic that's gaining a lot of traction right now is multiplexed or highly multiplexed immunohistochemistry. Immunohistochemistry is a 50-year-old technique of looking for specific proteins within a piece of tissue, where it's literally on the piece of tissue. So, I have not lost context of place. Historically, we're limited to looking at one to three different markers within that context. Well, now we have tools that allow us to look at 10, 20, 50, 60, 100, or even thousands of markers within the tissue itself without losing track of where they are in place. And now we need a standardized way of describing locations, so we can make use of that information without throwing the location piece away. And that's what we're talking about.

CONOR: And let's break this all down. remind ourselves again, histochemistry. So histo-, what's the root of that prefix? Where does it come from?

PAUL: It comes from tissue. Histology is the study of tissues, and chemistry refers to a study of the chemicals in context of their location, or the tissue itself. And so to go way back to the beginning of the invention of microscopy, van Leeuwenhoek was a very important early adopter, an early creator of microscope systems. And he first made little glass ball lenses and started looking at small things. He realized a lot of things were hard to see, he was in fact a textile merchant, he sold fabric.

And so he took the dyes that were used for fabrics, and he applied them to his biological specimens. And he could see structures that he couldn't see before. And as biology has adopted the use of looking at tiny things, every part of your body is made up of little tiny units called cells. If I want to study the cells, and I want to understand them, I need to somehow decorate them in a way that allows me to see them. That is histochemistry.

DODI: And, you know how one good idea leads to another? After talking to Paul, he said, “Hey, talk to somebody who knows about the science of mapping itself.” And so we went to find a professor in Indiana.

GUEST: My name is Katy Börner, I am working at Indiana University. I'm a professor here, and I teach data visualization. I'm also the curator of the mapping science exhibit, which has gone to many public places, to libraries, to science museums to National Academies around the globe over the last 16 years. It's a long-term effort, it's actually envisioned as a 30-year effort. It feels so lucky to sustain it so long.

DODI: Katy's written a few books, among them is one called Atlas of Knowledge: Anyone Can Map. And she loves maps.

KATY: Most maps are beautiful and actionable maps that help us find our way to the next party or meeting. They help us understand where we are in the context of the environment, but also in the context of political events, in the context of what can be done from a certain place at a certain moment in time. Maps evolve over time. And so I think many people are very, very curious to look at maps and to understand them better to use them intelligently in their personal and professional life.

And so we have to use maps as a way for different audiences to realize that there's now a lot of data available, to realize that there are tools that they themselves can use to analyze and visualize their own data, and to ultimately create their own map, be it of their own personal travels over geospatial landscapes or over intellectual landscapes. For instance, you might want to map your career trajectory of a landscape of science, but also to understand who has claimed what intellectual property if you are a lawyer, or to understand what educational opportunities exist, and how they align or do not align to this, what industry ultimately needs to hire for them, and other maps that you will also find in the mapping science exhibit. They look at Twitter™ data, they look at news data, they look at patterns and publications and clinical trials, data sets. And they try to help many understand what exists, how it's organized, and how many people can benefit from it.

CONOR: So, why does Katy think it's important to make mapping a science?

KATY: Many believe today that making a data visualization is an art form. And I think we can go much further than that. In my research, we try to understand how to measure and improve data visualization literacy, that has different types of expertise that make up data visualization, literacy, and that can also be tested.

And then there's a process of how these data visualizations are typically designed by picking a data set, analyzing that data set, visualizing the data set, and then ultimately interpreting what you see. And realizing, oh, my God, I want to zoom in, do I need more detailed data? Or I missed an entire year where I have to go back and clean up my data or get new data? Or to say, oh, now that I know what I have just seen, I have a new question. And this is actually the best type of data visualization. It is a visualization that helps you ask completely new questions, and hopefully also find answers to those questions.

DODI: I also talked to Katy about pet peeves when mapping is done poorly. And if there's real danger in that.

KATY: Anyone can map anyone, just like everyone can learn how to read and write or do basic math. I think today, it's very important that anyone really understands that they can take data from their personal or professional lives, and make a graph from it, or a geospatial map or network layout, or some trend prediction, etc. And just getting that into everybody's hands is challenging, but also a true opportunity for data visualization researchers, who ultimately are performing research in that area. And many of us are very excited to empower others to also create their own data visualizations. Ultimately, if you do this, then of course, they're also dangerous just like you can use a hammer to do many good things to build a houses and pots of houses, etc.

Many of the tools we have created for ourselves can be always used for good or you can also use them for bad. And similarly data visualization can be and has been used to communicate wrong data in a professional way, which then makes many believe that this professionally looking data visualization must be correct and true. And it might not be because the data that went into it was maybe coming from a source that is not a high-quality source or was even tampered with.

Ultimately, in the classes I teach, I help students understand that they now have a new superpower, they can now take data in ways that makes that data actionable for many. But with this power also comes responsibility. The responsibility includes that you should always document where the data came from, and you should always document the workflow that was used to get from the data to insight and ultimately, to make these data visualizations available in a way that many can benefit. But I think it can be done.

CONOR: And for Katy, it's better to map than not to map as long as you follow some rules guided by science.

DODI: That's right.

CONOR: Okay, I just want to back up just a little bit, because earlier Katy mentioned that with the superpower comes great responsibility.

DODI: She did, and of course that rings a little bit familiar to fans of Spiderman, right?

CONOR: Okay, she's a superhero. I love this that we have a mapping superhero.

KATY: I think many people have different superpowers. And it's a really important thing for anyone to understand what your superpowers are. I think in today's information age, having the ability to take on large-scale data sets, to master the many different algorithms that now exist, and to master the infrastructures that help us scale up computational processes to visually render this in an easy-to-understand format, that's definitely a superpower.

CONOR: So, it's safe to say that Katy is a real map expert for sure. Okay, so we can map careers and knowledge and information. Paul is trying to map tissues and tumors, which to me sounds like perhaps one of the hardest things to map. But what does Katy think is actually the hardest thing to map?

KATY: I think it's hard to map if you don't have data. It's hard to map if you don't have good questions about the data. It's also hard to map if you don't have the algorithms that scale to the amount of data that now exists. And in some areas, this really requires scaling up. Many of the algorithms exist now. But in many, many cases, you don't just want a static map, you want an interactive map that you can explore, can zoom in to search, filter, and get details on demand.

DODI: So, before we let Katie off the hook, so to speak, I asked her if she had any homework for you, our dear listeners of Discovery Matters.

KATY: I would send them to the mapping science exhibit website , because they will see 100 maps, 100 different ways to see our world. These maps were created by 240+ mapmakers from many different continents and countries. So, you might even get to see a colleague of yours, or somebody who speaks your own language, to really get a better understanding of this world to make more informed decisions, get us all to a more desirable future.

CONOR: I love it. I absolutely love it. Love it. The thing with maps now is that they're so accessible. It's absolutely amazing. I mean, you know, I've got a map of the whole world in my pocket all the time. Yeah, if I'd said that to myself at age 18, you're going to have a map of the whole world in one meter or maybe five meter photographic detail that would just blow your mind.

But here's my little problem with Google Maps™ and GPS: we don't always know where we are anymore. Because this is stuff that we don't have to remember. And it was brought home to me by a friend who drove to where we'd moved to our new house. He arrived, and he got out the car. And he said, “So where are we?” And I was like, “What do you mean, where are we? You just drove here, so you must have used a map.” And he goes, “Well, I used the sat nav, and I have no idea where I am.” And he literally didn't know where he was.

DODI: Yeah, because we look down instead of looking up. And the irony is, maps are supposed to help us look up and see.

CONOR: This is a devolution of our knowledge, because we're kind of outsourcing it to the digital world. So, I think paper maps are super important. It's a basic life skill to read a paper map, isn't it?

DODI: But they are for geography, Conor. Paul was talking about tissues and tumors, and you're not gonna have a paper map of that. And that's something that's going to be moving all the time, because these are masses that change constantly. So, they have to go digital.

CONOR: Absolutely. And that's where the power of computing comes in and totally revolutionizes the idea of what a map is. I mean, to have a map that actually changes as the geography, in inverted commas of whatever it is that you're looking at changes. I mean, that's absolutely fantastic. I'd love to look back on this episode in 20 years’ time and say, we would never have imagined that we'd have a real-time live feed of the whole planet in our pocket at any moment. Wouldn't that be extraordinary? I'm sure we'll get there.

DODI: And here's a big circle comment: One of the pleasures of discovering new things, or new places is kind of getting lost in them.

CONOR: Absolutely.

DODI: And maps are kind of a way not to get lost. But of course, you can get lost in maps, and you can get lost in the learning about them.

CONOR: Yeah, and that brings me down to my Sunday afternoons when I hear in the background, my children saying, “What's daddy doing? He's looking at his maps again.” I've been there with my atlas for hours poring over the Argentine plains or the Panama Canal or wherever it might be. Isn't the world an amazing place?

DODI: Well, we hope that you have enjoyed both getting lost and getting free in this episode of Discovery Matters. Thank you very much.

CONOR: And rate us.

DODI: Our executive producer is Andrea Kilin. Discovery Matters is produced in collaboration with Soundtelling. Production and music by Thomas Henley.

Listen to more podcast episodes.