For my story about the Nature Methods method of the year--spatially resolved transcriptomics--I interviewed many scientists.
In continued slow-pokey fashion, I am producing podcasts about this spatial subject. Here is episode two in a rolling series.
This episode is about smoothies, fruit salads and fruit tarts, about brain puzzles, about atlas-building and about the role of space in biology. It's with Dr. Hongkui Zeng who is director of the Allen Institute for Brain Science and Dr. Bosiljka Tasic who directs Molecular Genetics and her research is for example on cell types in the mouse brain, also from the Allen Institute for Brain Science.
It's about spatially resolved transcriptomics, which is a way to see where things happen in tissues. It's with two scientists:
You can also find this podcast on
Apple podcasts, Google podcasts, Spotify and wherever else you get your podcasts. It's part of a series called Conversations with scientists.
You can find more about spatial transcriptomics in the Nature Methods Focus issue. And here is a review on spatial transcriptomics in Nature.
Hongkui Zeng (left) directs the Allen Institute for Brain Science and Bosiljka Tasic directs Molecular Genetics there. (Photos: Allen Institute for Brain Science)
Transcript of podcast
Note: These podcasts are produced to be heard. If you can, please tune in. Transcripts are generated using speech recognition software and there’s a human editor. A transcript may contain errors. Please check the corresponding audio before quoting.
Not lost in space Episode 2
Hi and welcome to Conversations with scientists, I’m Vivien Marx. This podcast is about space--space in biology, actually. Talking about the role of space and spatial analysis in biology is a chat about food. About smoothies, fruit salads and fruit tarts. Here’s Dr. Hongkui Zeng and Dr. Bosiljka Tasic from the Allen Institute for Brain Science.
[0:30] Bosiljka Tasic
Fruit salad and smoothie. Fruit tart is spatial transcriptomics. Smoothie is Bulk RNA-seq. Ok passé
You have fruit salad, you have dissociated cells you are profiling, you have lost the context, you have a context in the piece of tissue you have dissected. Then there is the fruit tart. You know exactly where each piece of fruit in relationship to the other
Ok so spatial analysis in genomics is understanding a fruit tart. Knowing which genes are expressed where and what the relationship is of the genes to one another. The two scientists will talk more about this shortly. There’s Dr. Bosiljka Tasic, she directs Molecular Genetics and her research is for example on cell types in the mouse brain. And Dr. Hongkui Zeng who is director of the Allen Institute for Brain Science. Before they explain more about this science, here they both are, kindly teaching me how to pronounce their names. As ever I will try to do this right. And likely fail.
[1:37] Bosiljka Tasic and Hongkui Zeng
I'm Bosiljka Tasic. Bosiljka Tasic. OK, got it Hongkui Zeng. You don't pronounce the G at all, just, well, Zen, yeah, Zen G Zen. Yeah, yeah. It's very, very almost not there. How would you how would you pronounce that if you emphasize the G . ZengG. So I think g you hear much more but it's not the correct way. I mean I've given you my Americanized way of saying my name. I see. Well I'm going to, I'm going to do it wrong anyway. But but at least for me, don’t worry.
Next, before we get back to their thoughts and research, just a bit about this podcast series. In my reporting I speak with scientists around the world and this podcast is a way to share more of what I find out. This podcast takes you into the science and it’s about the people doing the science. You can find some of my work for example in Nature journals that are part of the Nature Portfolio. That’s where you find studies by working scientists and those are about the latest aspects of their research. And a number of these journals offer science journalism. These are pieces by science journalists like me.
This podcast episode about space in biology harkens back to interviews I did months ago. Back then I asked scientists about their work and their thoughts about spatially resolved transcriptomics, which is a Nature Methods method of the year. In my slow pokey DIY podcast production this is episode 2 in a series about this field of study.
Spatially resolved transcriptomics helps with studying the brain, which is the giant puzzle that Hongkui Zeng and Bosiljka Tasic work on. Among their daily puzzles is: How many different cell types are there in the brains of mammals such as mice, primates or humans? There are lots of them. And scientists want to be more precise than just saying there are lots of cells, of course. They want to know which ones there are and where they are. In the brain, another puzzle is where are cell types when. Cells are born and then often move to other areas of the brain where they will tend to all sorts of tasks. It takes a number of techniques to address these questions, including spatial techniques.
The US National Institutes of Health—NIH--has many research projects, one of them is the Brain Initiative, NIH's Brain Research through Advancing Innovative Neurotechnologies Initiative. Part of that is the NIH Brain Initiative Cell Census Network (BICCN).
One big BICCN project is to build a high quality atlas of cell types in the entire mouse brain. Many labs are working together to produce human, mouse and non-human primate brain atlases, these are intended as references for labs around the world. The scientists use imaging, electrophysiology and molecular genetic analyses including analysis of gene expression, which is transcriptomics.
BICCN phase 1 is underway and phase 2 is getting underway. The project has started with the mouse brain and is moving toward an atlas of the non-human primate brain and the human brain. One big challenge in this venture is distinguishing cell types. Cells may look very different but they might also look quite similar to one another. Here is Hongkui Zeng talking about BICCN
[5:20] Hongkui Zeng
We are currently in phase one, BICCN phase one, building this brain-wide cell type reference atlas. We are doing quite well and we expect to complete phase 1 in the next two years. And then phase 2 is starting, BICCN, phase 2 what you heard at SfN. There are several major themes for phase 2 that were announced by NIH.
The three major themes are building cell-type targeting tools, moving into the study of primate brains including human brain, cataloging cell types in the human brain and then finally studying the connections, the connectomics of the human brain. Bosiljka is very active in one of those initiatives, which is building in one of cell type targeting tools
You want to define a cell type first, but then you want to be able to access it for experimental examination perturbation. You want to form causality connections between a cell type and, let's say a specific behavior. So in order to do that, you need to build usually a genetic tool that is based on genes that are expressed in the cell type or maybe regulatory elements, enhancers that are active in that cell type.
You can you can create a transgenic mouse or a viral tool that will then deliver a particular transgene, a particular perturbing or labeling gene to that cell, and then you can visualize the cell, monitor it, maybe monitor its activity or perturb it and ask for Phenotypes effects at the level of that cell, at the level of the circuit, at the level of the whole organism.
And both Hongkui and I, we are we have a just accidentally sort of independent histories and building genetic tools. And then at the Allen, we sort of merged our forces, but both of us worked on building genetic tools. And then here we worked together on. Again, expanding and building new genetic tools, but for me, this is something that I've felt was always essential. You can define Cell types, you can define exactly where they are in the tissue, but you need to do something about them, right. To visualize them, but not only visualize them, you need to perturb them. And then you need to observe the effect that perturbation has on the organism. That's how you build causality.
Atlas-making and genetic tools in brain science are about analyzing cell types, knowing where they are in the brain, learning what the cells do, how they interact with other cells and how their activities lead to complex behavior such as memory. Part of this science undertaking is knowing which genes cells express where. Genes tune all sorts of things in the body and the brain, too. Genes might be turned off for a while, then be on and highly expressed. They might have low levels of expression or be silenced for some phases, expressions can shift. Knowing which genes are expressed where is at the core of spatial transcriptomics. Hongkui Zeng explains how spatial transcriptomics matters in brain development.
[8:57] Hongkui Zeng
So spatial transcriptomics is also critical for understanding development because during development, the number of cells, not only the number of cells is increasing. Right. And regions are growing, but also there are there is migration of cells, all kinds of cell types happening. And the and the cells migrating. They follow specific trajectories. Very often the cells migrate over their long distance from where they are born to their final location. So the state of the cells in development is often associated with the position of the cell during that path.
There's a lot of migration happening. Think about your whole brain or your body comes from a single cell. And then there are always new cells are born and they are all organized in this beautiful structure. Cells are moving during development all the time. So you trace their past, then you understand, you know, that kind of relationship across time.
Spatial techniques can yield plenty of valuable information about the brain for Hongkui Zeng, it started nearly 20 years ago. Allen Brain Atlas And over time it’s become clear it’s hard to distinguish cell types.
[10:28] Hongkui Zeng
For me, it started with the Allen Brain Atlas on a Brain Atlas started in two thousand three. It's a in situ expression profiling of all the genes in the mouse genome, about 20-25,000 thousand genes in the mouse genome. It look at it, it looks at the anatomical spatial expression patterns one gene at a time. It’s a reference database that has been widely used, extremely useful, and people have learned a lot about cells that express individual genes.
Over the years we have been using that atlas to try and understand the different types of cells in the brain, how many types there are, however. Well, along with that, you know, there are also additional techniques developed like double fluorescent in situ hybridization, triple fluorescent in situ hybridization, because we want to look at co- expression of genes Very often one gene is not sufficient to identify a cell type.
We were acutely aware that of knowing just one genes, two genes or even three, the combination of genes is not sufficient to identify a cell or cell type. So Single cell transcriptomics technique, you know, came in several years ago. It really changed the field revolutionized the field because you can look at the expression of thousands of genes in the same cell at the same time, and that's just tremendously powerful.
So that has already changed the field dramatically. And now we get into spatial transcriptomics with the different techniques that also allow you to look at maybe not thousands of genes yet, but, you know, it depends on the type of method that we're talking about. But it's the same idea that but go even beyond the single cell are rna-seq kind of technique. It allows you to look at several genes, what many to use in a same cell and spatially localized region. You know, at the same time, it just, yeah, very powerful when you identify the identities of the cells and also exactly where they are located and what cells are near their cells. Spatial organization of the cell types.
Now we can measure not one gene at a time, but you measure thousands of genes at a time, maybe not in the spatial transcriptomics context yet, but in single cell transcriptomics can measure thousands of genes at a time. And you measured their levels and you actually notice that definitely quantitative differences are highly prevalent. Actually, they're more than of a rule than an exception, meaning black and white frequently does not exist, You have multiple genes that are expressed at multiple different levels and together they make a cell what it is.
Vivien Together the expressed genes make a cell what it is. When scientists analyze which genes are expressed in a tissue, they analyze the messenger RNA, which is actually a tiny fraction of the RNA in cells but a really important one. And as you already heard they would rather not have all the genes all blended together into a smoothie, but they would like to see where the individual genes are expressed and get that fruit tart-y view.
[14:08] Bosiljka Tasic
So I think a couple of things matter: how many genes do you want to detect at a time? How highly are these genes expressed? What is the how accurately you want them quantified? Do you really want to count molecules where you're OK with just counting overall signal? Do you really can you image in thin sections versus thick sections? Do you need to image in a volume or you're fine with sectioning your tissue very thinly? So I think those are the main things, did I forget anything Hongkui?
I think, yeah, in addition, there is also do you want to do you want to have high resolution, high cellular resolution, but only local looking at a local region, or do you want a overall survey across a large area, part of the brain? But you can tolerate resolution. You don't have to have single cell level resolution, some kind of a local, resolution is sufficient. In a former case, you would use MERFISH type of hybridization-based approach, in the latter case you can use 10X Visium, that kind of spatial transcriptomics.
You can capture a large area and it's very high throughput. You can look at many and it's very fast. You can look at the many, many cells from the same section simultaneously. But of course the sensitivity, you sacrifice sensitivity, you sacrifice single cell resolution but it’s high-throughput. Vivien With spatial transcriptomics many see its starting point with in situ hybridization that scientists applied to find a particular bit of DNA or RNA and get its location. Single molecule fluorescence in situ hybridization gives an enhanced signal for localizing a molecule such as an RNA.
[16:13] Hongkui Zeng
Single molecule FISH is the beginning where you can use multiple oligos to bind to the same molecule and use that to enhance the signal of the detection so that you can see Yeah, very low level single molecule. You can image single molecules from a tissue section. OK, that's at least that's the beginning of the hybridization-based approached Vivien: Many spatial transcriptomics methods can involve sequencing and those can be divvied up in different ways. Here is how Hongkui Zeng classifies them.
[17.00] Hongkui Zeng For me, it's really just three main approaches One is in situ hybridization, like MERFISH multiplexed, in situ hybridization. The second is In situ sequencing, and then the third one is In situ capturing, followed by bulk sequencing. So that's like the 10X Visium kind of approach. So in situ capturing, n situ sequencing and multiplexed FISH.
All of those technologies currently follow single cell RNA sequencing of isolated cells or nuclei because the single cell RNA sequencing from isolated cells or nuclei still give you the highest sensitivity in terms of detecting the number of genes expressed and things like that. W e use spatial transcriptomics either of those three approaches to do subsequent studies to look at where the cell types are distributed in the tissue.
Vivien One aspect that may be surprising to some is that spatial transcriptomics involves gene expression and that the methods currently involve sequencing technology. But maybe, maybe single-cell RNA sequencing is no longer needed to get the expression data needed in spatial transcriptomics. Hongkui Zeng explains.
[18:30] Hongkui Zeng
What we want to see ideally one day is that we can completely get rid of single cell RNA sequencing. We can just do, yeah, do it at once. If it really has sufficient sensitivity and a resolution and the ability of detecting thousands of genes, then you can just take a tissue infection and then just, just do a spatial transcriptome-wide.
You've got all the information. You don't have to do two steps. You just do it once. And then when you take, let's say, a human biopsy tissue or whatever you want to do, you can just you can just look everything, you know, and do it once. Got all the information, you'll be much more efficient. And you've got both spatial information and a quantitative gene expression information as well. Vivien: Although this might sound to some ears as if this is still a development that is quite far off in the future, Hongkui Zeng doesn’t think it’s that far off in the distance at all.
[19.35] Hongkui Zeng
It's maybe happening pretty soon. We count on technologists like Xiaowei or companies like 10X Genomics to develop those technologies that will allow us to do that. Vivien: One reason it matters to make things efficient with technology is that in plenty of instances, a lab is analyzing tissue and it’s a mighty precious sample to analyze. It might for example be a sample of a child’s brain after surgery for a brain tumor.
[20:03] Hongkui Zeng
There are samples, unique one of a kind sample that can only be examined once, you know, that kind of sample. Vivien In cases such as these, spatial transcriptomics has much to tell basic researchers and clinicians, who want to know more about a child’s brain tumor.
[20:20] Hongkui Zeng
There is huge diagnostic potential.
I think, diagnostic not only definition of cell types in healthy brain, but diagnostics is going to be revolutionized by single cell genomics techniques. It's pretty clear it's there. Just imagine, now we measure something that's in your blood, some protein. Now, imagine even on a tissue that's not spatially organized, you can sequence RNA in all those cells. And imagine in a tissue that's organized like a tumor, solid tumor imagine you just image all the genes and their expression in that tumor, you know, the spatial organization and you know what's special or different about these tumor cells in this case versus that other case.
You might spot that, you know, very rare tumor, precursor cell, you know, things like that. That's what you want to find.
Vivien These rare cells might be exactly where a tumor originated. Knowing this cell might be crucial to treating a patient and it might deliver to basic researchers an important clue about this tumor type and cancer. There are many methods labs can use to locate where genes are expressed to get spatially resolved transcriptomic data. Boslijka Tasic and Hongkui Zeng talk a little about what researchers need to consider when they decide what to use when. They will mention a few approaches that you will hear more about in other episodes in this podcast series. There’s MerFISH for example from the Lab of Xiaowei Zhuang at Harvard University and 10X Genomics’ instrument called Visium based on technology from multiple labs in Sweden. Episode 1 of this podcast series was with some of those developers.
[22:20] Bosiljka Tasic
Let's say you're interested in a particular cell type and you want to detect it robustly. The question is, is it from other cell types? How frequently is it present in a tissue? So both of those things, decide what sort of influence, how you will design your experiment if a cell type is very distinct and very abundant. You actually may not need spatial transcriptomics at all. You just need almost old-fashioned Allen Brain Atlas type of approach. But that's rare. A single gene defining a cell type is rare. Now, if we want to define finer and finer cell types that depend whose definition depends on more genes, we need to multiplex the genes. And we also Want to make sure we can distinguish that cell type from, let's say, related cell type.
Sometimes we even want to define a class or examine a class, a group of cell types that are really so you have to define where you want in resolution to be. Do you care about defining the finest divisions in the taxonomy? Define a cell types or you just want to look at a class like, for example, parvalbumin interneuron class. Based on that, you design the set of genes you will look at. And the last thing is also, do you want cellular resolution? This is going back to what Hongkui mentioned before. Do you really need to detect individual cells or you care about how abundant is a particular cell type in this area? And is it in a particular area if I have some cellular resolution, that's Visium.
I think the best thing is what everybody dreams about is I just want to measure all the genes in all cells with cellular resolution. Everybody would love to do that if it were cheap and easily accessible. But then there are many times where you just don't need to go that deep. And actually, currently we don't have a method that can measure all genes in every single cell, their expression at the cellular resolution, we have MERFISH which can measure a couple of hundred very accurately, and we can infer other genes, their expression based on, for example, a combination of single cell transcriptomics with spatial data, with MERFISH data.
On the other hand, sometimes I just want to look at three genes, because these three genes would already tell me the difference between the three major classes let's of inter neurons in the cortex. So I don't need to do spatial transcriptomics, but I know which genes I'm going to probe.
So that's the key thing. Discovery. We still do discovery mostly at the level of single cell RNA seq on dissociated cells. I think Xiaowei might slightly disagree with that because she does do clustering, her lab has pioneered clustering. So de novo cell type discovery on MERFISH data. But you still need to know which genes you are going to probe. You still have to have the information which genes? I'm going to choose to have the best possible coverage of the landscape of cell types, and you have to base it on something. You base it on single-cell RNA-seq.
An overarching dream is measuring all genes in all cells at single-cell resolution. Here’s Bosiljka Tasic.
[25:29] Bosiljka Tasic
When I was in grad school, I was dreaming of this. I want to measure all the genes in single cells and so many single cells And people were pioneering it even then. I mean, their early single cell transcriptomics type of work. And not to mention in spatial context, I mean, this was a dream for a very, very, very long time. Vivien The spatial dream has been a while in the making. And now that it’s a reality, people need to craft their experiments with a view to, for example, what kind of resolution they need. One issue with spatial analysis and the methods that exist : labs are getting a lot of data. And it’s not just the transcriptomic data.
[26:10] Bosiljka Tasic
We ourselves were not even aware how quickly, we will be hitting the boundaries of what is standard to be analyzed. So just imagine you're measuring with single cell transcriptomics, but similarly with spatial transcriptomics you're measuring each cell and hundreds or thousands of properties for each one of them. So that means, let's say, single cells, transcriptomics will have millions of cells measured for forty thousand genes.
In spatial transcriptomics, you will have maybe once we have the whole brain, all the cells in the brain measured for two hundred and fifty genes, just the scale of the data becomes and these so-called data matrices become unmanageable. So what we have done is we are developing some software internally, we try to repurpose and adopt other people's software, but in fact, we have realized we actually have to work with people who work with large data that has nothing to do with biology. They just work with managing and accessing and sampling large-scale data matrices.
We have to work with people who don't deal with biology, but we are now faced with this huge amounts of data. And how do you manage it? First how do you put it in one place and then how do you process it? So what we do is we do all sorts of things internally, develop software, collaborations, adaptations of other people's software in working with people who just work with large scale data, not biological data. Vivien Data mountains bring on all kinds of challenges, just their sheer size and the fact that the data come together from many different methods.
[28:03] Hongkui Zeng
When you deal with data, it's I think the first problem is the size of the data. The amount of data size of data just increases dramatically, very, very quickly, because the technologies allow you to collect many data rapidly for the second issue. And problem is, as you said, there are now multi modal data, many different types of data. And when you deal with many different types of data, you have to find a way to be able to integrate them together. Even the same type of data collected in different ways are somewhat, slightly different.
So you need to correct batch fact and donor effect. All these things find ways to normalize the data that are collected, the same type of data but collected in different ways. So that's one way of integration.
And then next, the integration is, as you said: there are sequencing datasets, there are imaging data sets, there are physiology data sets. And how do you combine them together integrate them together to perform joint clustering analysis, for example, or look at a cross correlation across different data sets. So both of those types of problems involve, you know, computational techniques, mathematicians. Nowadays, a lot of a machine learning, AI-based approaches are very powerful in integrating data together as well. Vivien As part of this integration, scientists are developing new ways of looking and analyzing their data. When it comes to data integration, you can’t just throw data together make a data smoothie.
[29:53] Hongkui Zeng
You tweak, you play with data, you play with different parameters and you develop new equations, algorithms that are specifically suited for your problem
It is really essential for biologists to work closely with bioinformaticians and mathematicians for progress in this area to be made. And as Hongkui mentioned, not only because we have large sets of data, but the dimensionality of the data is increasing and the modalities, the types of data we collect for sometimes for the same exact cell we can collect three different types of data
Vivien Within spatial transcriptomics there are many different ways to capture data about where genes are being expressed. This multitude of methods is true in fields that are rapidly evolving. And there is still a lot of room for improved approaches because for Hongkui Zeng, Bosijka Tasic and many of their colleagues in neuroscience, the brain’s complexity is an intense challenge.
[30:57] Hongkui Zeng
It's a very active field, as you have seen. There are so many different techniques that have been developed to really open up many opportunities for us to look at different things and think about new experiments that we can do. But even with that, despite of that, the techniques are not in many cases, They are not perfect yet. And we want we want to more and better.
There are many questions in our mind that we just that we can scale up that technique, you know, looking at not just hundreds of genes in a cell, maybe thousands of genes in the cell and being able to measure things much more efficiently. Vivien For some of the techniques in spatial transcriptomics scientists can turn to a commercialized instrument. There are many companies in this space including 10X Genomics, nanoString, BGI and other companies and new companies are emerging.
[32:00] Bosiljka Tasic
Commercialization does help. Having some of these because academia, I mean, you develop methods and you use them, but in fact to make them widely available, commercialization sometimes is really crucial. For next generation sequencing, for single cells transcriptomics without the next generation sequencing, this revolution in single cell transcriptomics wouldn't happen.
And I think. I'm very much looking forward to commercial solution for spatial transcriptomics that will work at the Single cell level. Instead of sequencing RNA that I isolated from individual cells, can you just sequence things directly or imaged them on a piece of tissue and give me back the data.
Many, many of the kind of work that we do we're doing it for the first time. When you do something for the first time, it's always difficult and expensive. However, once you are able to do it and then you develop a technology that allows you to do that repeatedly, many times that's that's when that's the way the applications will take place. Vivien For now, spatial transcriptomic analysis is not routine. Bosiljka Tasic likes to imagine a day when routine sets in.
[33:12] Bosiljka Tasic
I think it will become as standard as Sanger sequencing. I think it will be a standard is Sanger sequencing, you will take a piece of your tissue and you will send it out at some point and you will get a spatial transcriptomics image or your tissue would sell a resolution or not, depending on how much you pay, which technique you choose. I can see it is becoming a standard thing that even in papers. Oh, you didn't measure all the genes? Why not? Vivien Today and in the future not every lab will need to know the gene expression in every single cell in their sample.
[33:50] Bosiljka Tasic
I mean, for many, many purposes, that is also still infinitely useful. If you can measure a hundred or two hundred and sometimes even just four or five or six is really useful. But I think one can imagine that one day you won't be doing single cell or any sequencing, but disassociating cells at all, you will just do spatial transcriptomics directly. Vivien Spatial transcriptomics and other techniques will help with atlas building of the brain. At the Allen Institute researchers are building atlases of the brain to know and show which cell types are where. The idea with such atlases is to give labs around the world the opportunity to compare their results with these atlases.
[34:35] Bosiljka Tasic
A major thing really is what we call cell type annotation cell type mapping. How do you define within your sample? What are the cell types and can you infer what they are based on somebody else's data? And in fact, this is this is really something that many, many people are thinking about, standards for defining cell types, standards for atlas's of cell types with cell type cards and ways to map your cells that you have sequenced either in your lab or through a service to that standardized taxonomy of cell types so that you can see this tumor has this much glial glioblastoma cells versus this tumor has medulloblastoma cells, the proportion of different cell types and their spatial organization based on comparison to a standard set like a periodic system of cell types.
We're very much looking into building, I mean, a periodic system of cell types in the brain. How is it exactly going to look like? It's not going to be periodic. We would like it to be. It's probably going to be quite complicated, but we would like to define every cell type provided to the community as a resource and enable the community to map their cell types or their cells to our standardized cells. And also, I should mention, it's not only us, it's not only the Allen Institute we're part of large consortia
Brain Initiative Cell Census Network and Hongkui is a one of the main leaders there.
Vivien In neuroscience there are still plenty of disagreements between scientists about which regions and sub regions of the brain are responsible for what.
[36:25] Bosiljka Tasic
That’s because we're still working on the cartography once with the cartography is complete, then I don’t think there will be a lot of disagreements. But then there will be the questions of function. Right of cell types. What do they do? How do you perturb this one or that one? How specifically do perturb cell type A or B, how did you measure the effect? There will always be disagreements in science, but I think with time and with a lot of data collected, they also.
Vivien And then there are disagreements still about spatial data itself.
[37:02] Bosiljka Tasic
They occur because people use somewhat different methods, because execution of these methods in obtaining these data is difficult. It's not because people really want to disagree, but because it's really hard to get these data and then to interpret them in a joint way. It's also hard. So that's one of the reasons why I'm a huge fan of common coordinated system, for example, for the mouse brain where you put everything in the same space.
That's another reason why I like transcriptomics, because you measure the same genes. You can really correlate data from lab to lab, you can integrated science so far has been done frequently in a way that you create data and you publish your paper and it's a 2D image in a publication and nobody else can use it ever again for anything.
Vivien At the Allen Institute, the scientists want to provide atlases and a way to enable better ways to standardize of data and make it easier to use and compare. And yes, there will still be papers in the future. Bosiljka Tasic and Hongkui Zeng explain.
[38:12] Bosiljka Tasic
We’ll still have publications. Public data deposition in a standardized format such that other people can use and reuse those data, compare it to their own, it's essential. Genomics has been moving in that direction for a long time, I think image analysis is also moving in that direction. Hongkui, for example, led a Connectivity Atlas at the Allen Institute, where all the experiments were placed in this one common coordinate framework. Allen Institute Common Coordinate Framework, is basically a standard brain for the mouse. So you don't present data one at a time. You present them all in this one common framework, and then you provide a common framework for everybody else in the world to be able to use it. Vivien This kind of atlas needs to be built and labeled.
[39:05] Bosiljka Tasic
3D map of a mouse brain with areas annotated and labeled with names
[39:15] Hongkui Zeng
It's a reference 3D space. So then everybody can register, map their own data into the common 3D space space to allow comparison, objective comparison. Vivien These activities of cataloging are about what is here. As it turns out atlas building involves new discoveries, too.
[39:30] Bosiljka Tasic
We have it we have everything has to be version because also now as we're sequencing RNAs in these individual cells, we're finding new mRNA isoforms. That means we are also finding new genes that haven't been seen. So that's why I keep saying we said 20,000. That's roughly the number of genes in the mouse genome But it's probably maybe fifty mostly.
Mostly coding, protein-coding genes. If we count new non-coding RNA species that have been identified, then it's a lot more. Vivien Now after you have heard a bit about various methods in spatial transcriptomics, let me re-share the comments of Bosiljka Tasic and Hongkui Zeng about smoothies, fruit salad and fruit tarts.
[ 40:21] Hongkui Zeng
Fruit salad and smoothie.
Fruit tart is spatial transcriptomics. Smoothie, Bulk RNA-seq. Ok passe
You have fruit salad, you have dissociated all the cells you are profiling, you have lost the context, you have a context in the piece of tissue you have dissected. Then there is the fruit tart. You know
That was conversations with scientists. Today's episode was with Dr. Hongkui Zeng and Dr. Bosiljka Tasic from the Allen Institute for Brain Science. And I just wanted to add, because there's confusion about these things sometimes, these scientists and their institution did not pay to be in this podcast. This is independent journalism produced by me in my living room. I'm Vivien Marx. Thanks for listening.
This is from NASA Galaxy Evolution Explorer and shows Andromeda galaxy, or M31. It's around 2.5 million light-years away from the Milky Way. (Image: NASA/JPL-Caltech)