Today I talk to Viktoria Korzhova, a PhD researcher turned into a science communicator and manager. Viktoria is working towards a PhD degree from Systemic Neurosciences school at LMU Munich, and recently embarked science communication and management path.
Today we discuss how technologies in imaging and data science help to study neurons. We explore what is the routine of a researcher, how it can be improved via tech and how these technologies can be repurposed. Viktoria also shares some ideas on building a career in science and discuss collaboration with the research community.
The talk is a comprehensive one and is organised in seven parts:
Part I. Basic concepts: animal brain imaging, Alzheimer’s disease
Part II. In vivo neuroscience experiments, exploring single-neuron level
Part III. Exploring neurons in the human brain, fMRI, modeling of the whole brain, simulations
Part IV. The routine of studying the animal brain: labelling and analytics, related software
Part V. How tech may improve brain research?
Part VI. Repurposing tech within the lab
Part VII. Science and entrepreneurs, a research career .
Part I. Basic concepts: animal brain imaging, Alzheimer’s disease
Peter (P): Today we talk with Viktoria, whose experience is in quite interesting field of brain imaging. And the starting point of our talk is the introduction. Can you please tell us about your educational experience and what are you doing?
Viktoria (V): First, I’d like to make sure we use the same terminology. The brain imaging we’re going to talk today about is not the human brain imaging. It’s about imaging of the animal brain and what is also very different from human imaging is that we image at the very different level. So in my research I imaged at the level of single cells, single neurons, whereas in human brain imaging we always image brain more or less as a whole or as some selected area which then includes hundreds of neurons and we basically cannot have resolution of single neuron level.
I’m a biologist by training, I have graduated from Saint Petersburg State University with a masters in biology and my specialization was physiology and genetics. Then four years ago I started my Ph.D. studies in the field of neurobiology and I worked on the Alzheimer’s disease. This is currently the most common cause of elderly dementia and possesses a big problem to our society and the healthcare system because we don’t have any good treatment and the treatment we have is helping for only a short period of time and we don’t have any cure and even worse we don’t even have a clue where the cure can come from.
In my research project I studied the specifics of this disease on the animal model, on the mouse model of Alzheimer’s disease.
People ask me how mice get Alzheimer’s or do they get it at all. They do. But this is because of humans, so we humans we make them get Alzheimer’s because these special mice are genetically modified. So normally in wildlife mice don’t have this problem but because we genetically modify them in the specific way they also develop symptoms of Alzheimer’s including the special protein aggregates that we can’t find in the brains of affected patients. This is also what we find in the brain of these mice. And also these mice get some cognitive impairments because that is kind of tricky to judge if mice are cognitively fine or not but they do have some memory issues. And my task in this project was to study what happens with the activity of neurons, how neurons in the brain of mice behave when the disease starts developing. So when the early stages of the pathology happen.
P: How did you find yourself in this field in particular? As far as I understood from some of your previous posts degenerative diseases were not the first choice for you. What was interesting for you was the methods behind. Am I right or it’s not as simple as that?
V: I think it’s never simple, there is a combination. So on one hand when I started at the university I was interested mainly in genetics so I thought it is super cool to study how genetics affects our life in so many ways and basically determines so much about our health and our behaviour and I was interested to work somewhere on the intersection of behavioral neurobiology and genetics. But I actually ended up in the molecular biology lab where we studied prions and prions are a special kind of infectious agent which are very unusual because these are not bacteria nor viruses as most of other infectious agents are but prions are protein particles and these particles can infect living organisms including humans and can cause a brain disease. So by this I got an introduction into the field of neurodegenerative diseases, and prion diseases have something in common with other neurodegenerative diseases like Alzheimer’s or Parkinson’s or Huntington’s disease.
And from then on I actually for the rest of my research career worked in the field of neurodegeneration. Just changing a level I work at. So I started to work at the molecular level and then I worked more at the level of single neurons living in a dish. Then I moved to work with a living brain. So that’s what I was doing during my Ph.D. training — imaging the brain ‘in vivo’ meaning inside of a living mouse.
Talking about methods, I was interested to get to know different methods’ perspectives on the subject. Also I think that the method of in vivo imaging is something really exciting and is one of the very promising methods in the current neurobiology if we talk about animal research.
Part II. In vivo neuroscience experiments, exploring single-neuron level
P: I think it’s very exciting point, this in vivo method, when you collect data or make observations from living species. And I would say that we already see it in tech/data science, when historical data is enhanced by data from sensors when you can track a real actions of a customer, for example. That is one of the most exciting things, when you try to understand what is your customer doing based on the data from sensors. So for me this is pretty much the same concept when you’re trying to track something which is active.
Can you describe some routine, what does it mean to study an animal? How does it work? So you get these genetically modified mice and what happens next?
V: Yes, so of course there are many ways you can use this genetically modified mice and you can do not only in vivo imaging. You can also make other experiments, for example, can study the biochemistry and not only of the brain but also of other parts of the organism to see if they are somehow affected by the Alzheimer’s disease.
In our study, I need to expose the brain cells to the outside world, so that I could look with a microscope at these cells. This meant that I would be making a hole in the skull of a mouse and covering this hole with a piece of glass. So that the brain was intact and it was still under the cover. This glass enables scientists to look with a microscope directly at the brain because you cannot do this through the skull as the bone will not allow you to look with a microscope.
So this is one part, and the second part which I also needed to do is to label neurons with a specific molecule. Normally when you look at neurons, they do not show anything exciting. They’re not moving, they are not like muscle cells that contract when they are active. Neurons don’t show anything when they are active.
So I need to do something to see when they are active and when they’re not active. To do that we use a special type of compounds which are called calcium indicators. And calcium is a small ion that is used by the neuron when it is active.
So when the neuron is active there is increased concentration of calcium inside of the neuron. When the neuron is not active, or we say “silent”, there is decreased concentration of calcium inside the neuron. So the calcium concentration is coordinated with peaks of neuron activity. The principle behind these calcium indicators is that there is the specific molecule which can emit fluorescence. So when we light the specific wavelength on this molecule it will light back in the specific wavelength. This is called fluorescence and these molecules can only be fluorescent when they bind calcium and when there is no calcium and they cannot bind anything then they are just invisible.
And this is exactly what allowed us to see activity of neurons. So we’ll look with a microscope onto these neurons and we apply a specific wavelength which is suitable for that calcium indicator and when the neurons get active then there is an increase of calcium in the inside of the neuron and there is a peak of fluorescence of the calcium indicator and we can observe this visually.
P: Is it correct that you can use different wavelengths for different parts the brain and they kind of flesh out different colours to say simply? Or it’s not like that and they all have the similar kind of light.
V: The wavelength is the attribute of the calcium indicator. So dependending on which specific molecule you use, you can have different colours.
P: Why we don’t use MRI or something else that does not require to drill a hole?
V: So it of course depends on what you want to study. With fMRI we can only study the ensemble of neurons, so we can say that this ensemble of maybe hundreds of neurons behaves in this way. But we don’t know how each of these neurons behaves. If we want to understand how the molecular changes affect the functioning of a single neuron, then you need to look at the single neuron level. We need to have better resolution basically. Metaphorically speaking, if you look from outside at the forest then you’ll just see a lot of green and you can say okay the forest is all green. But if you want to know if different kinds of trees experience different changes throughout the seasons you need to come closer and you need to distinguish different tree types and how they behave.
P: And let’s imagine we try to do something like this with humans… drilling holes… are we doing that right now?
V: We normally don’t. So with humans we don’t currently do something like that for several reasons. First, we don’t want to drill a hole in a human’s head. And human experiments are not accepted in most of the civilized world. Second reason is that this calcium indicators that we need to use for that they are not completely safe. I mean they are fine, cells don’t die because of them but they still alter a little bit how cells are functioning. So in animals of course we are ready to pay this price because there is no better way for now to study the activity of neurons of individual neurons. But in humans we don’t want to do any damage.
Part III. Exploring neurons in the human brain, fMRI, modeling of the whole brain, simulations
P: And let’s just spend maybe a couple of minutes and speculate how it may work for humans. I mean, not drilling, but let’s imagine that we have already studied the brain of mice or maybe larger animals quite well and answered many questions there. Is there any chance that we can look at the human brain through non-invasive methods like fMRI and then back-solve using the data and knowledge we have from animals. Does it make sense, or it doesn’t work like that?
V: I think we will still need to go from two sides so on one hand we need to study how the neuron functions and we need first to understand how the human neuron is different from mice neuron. And this is something we don’t have a clear understanding yet.
To understand human neurons better we can study human neurons in a dish. For example, one situation when it happens is that when a human has to undergo a brain surgery. Sometimes we can get that piece of the brain. This happens with epileptic patients.
There is basically a part of the brain which behaves abnormally and in case the medication doesn’t help the best solution sometimes is to remove this part. It’s not a big part, it’s a small piece of the brain. But then this brain can be given to researchers if the person agrees and then the researchers can keep it alive for some time and study what is special about these neurons. And this is one way how we can get more insights about the specifics of certain human neurons and then we can compare proteins which are inside of human neurons and mouse neurons and how they activity is different.
Another way to work with neurons is to get the neurons from the pluripotent cells, those that have several different possibilities for development. We can find these in embryos: when an organism starts developing it basically grows from one cell, cells divide and then they start to specialize and become specific cells on the brain of blood of muscles and so on. And there is a certain stage at which cells are still not decided and can go in many ways. We can grow these cells in a dish and send specific signals to them so that they would became one or another type of cells.
There are two major sources how to receive this cells. One is from the aborted embryos which is not easy because in many countries it is not allowed to use embryos. But luckily in the recent years we got a better solution. These are so-called induced pluripotent stem cells. We can get these cells by taking an already specialized cell from the human body for example a cell of the skin. We apply special signals so that the cell will go kind of back in its development to become again a pluripotent cell and then we can make any cell out of it.
So this was a long introduction for me to say that on the one hand we still need to study the single neurons of the human and understand better what’s special about how they function and if we can relate so well the animal neurons to human humans.We should also do studies on the whole brain level, with fMRI, for example and then we need to do the similar experiments with animals and see what’s the relationship between that activity of a single neuron and neurons of the whole brain, or a specific area.
And here, of course, computation plays a huge role. One of the things scientists try to do is they try to take their mathematical description of a single neuron and then model how would the ensemble of neurons behave or how would hundreds of these neurons behave and how the whole brain behaves. Then we can compare it to experiments and see what is matching and what is not matching to make the model better.
P: So I assume there is demand on software that allows to build huge simulations?
V: Yes, certainly. There are already people working on that. It’s already obvious that with current technologies you’ll need a lot of computational power so modelling the whole brain is really far from current tech. There is a big project dedicated to the modelling of the brain it’s called Blue Brain Project. It’s a big European project which tries to figure out how we can do that and what are also the parameters that you need to use to have this model represent a reality.
Part IV. The routine of studying the animal brain: labeling and analytics, related software
P: Let’s look at something that is in our reach, something that is already happening. So you have the mouse with the whole in its head, what’s happened next, like you observe it every day, every hour? How does your routine look like?
V: In my case I observe the activity of neurons every week. So once a week I take mouse out, bring it to the microscope and let it sit there for some time and then I do recordings. Of course I don’t just look with my eyes what has happened, but I record the fluorescence of that specific area that I study with the camera and my recordings last for approximately 10 minutes.
I tried different setups of the experiment and what the final set up of my experiment includes is five measurements. I measured the activity of the same neurons because I also can find the same neurons next week.And then the question was will this activity change somehow over this month. Another angle is to check if this activity is different from the activity of neurons in a healthy mouse, or a mouse that doesn’t have Alzheimer’s.
P: How does your work with these recordings look like? So you record the video and then you somehow label its content or you write some narrative and describe what you have seen?
V: What needs to be done with this video is first I need to mark the areas which present neurons. And this is already not an easy task and tech did not get us very far with this. Recognition of cells in biology is quite a tricky task. There is software developed that does it automatically. This only works for a limited number of applications and in my case, for example, I had to label neurone semi-automatically.
First challenge is labelling of the same neurons through time and several observations. I label neurons manually and then I can transfer this labels to the next observation, just checking that labelling is correct. But there’s no good algorithm that can reliably detect neurons in the type of recordings that I do. An alternative approach would be that you ask an algorithm to detect neurons but still you need to check all of them and often edit labelling. So this is the first part and its purpose it to mark areas that will be analyzed for fluorescence intensity by another algorithm.
This algorithm looks at each of the frames of the video and gives it a number which means how bright this area was at this frame. That helps to spot changes in the level of lighting, you get a graph that oscillates because the neurons get active and non-active all the time, they switch the mode fast and these bright sparks are quite short. And this is also the reason why we actually do the recording, for an eye it might be very difficult to spot and count it. So each of these bright peaks is called spike, the peak of neural activity.
The next part of the data processing is to detect how many spikes happened during a time period and then calculate what’s the frequency of the activity of this neuron.
And here though it seems a very clear tasks, actually there are many tricky things, as you need to specify the threshold of what is the spike. The data is never super-clean, it’s always a bit noisy because it’s a living animal. And also hardware we use to record the video is very sensitive and it can detect some activity that actually doesn’t mean anything. And then because this is electronics there always will be some noise as well. And the scientist need to decide what will be the threshold below which it doesn’t matter and above which these are the real spikes. And then you get the frequency representing how active this neuron was and then you compare the frequency of each neuron between these different time points.
One can also divide neurons in different categories e.g. based on their activity level, patterns of activity, group those that are active together, etc.
P: Are you doing it manually by looking into a spreadsheet, or there is software that helps?
V: This is mostly done with the help of software but it’s not completely automatic. Firstly, there’s no software which is sold in the box that you can use for this. And there are not so many researchers which work with that and then each of them has more or less their own script to analyze that. In my case I worked with a collaborator who is versed in Matlab and understands specific data that I worked with, so she could write code to analyze my data. But of course I needed to tell her what actually I want to measure. For example, if I want to group neurons what are parameters I want to use. As neuronal activity is normally distributed there is no clear cut-off that these are super active and these are medium active. And of course the cut-offs researchers choose are partially arbitrary. I selected the threshold relying on published research. And this made possible to compare my work with other studies.
Part V. How tech may improve brain research?
P: So I see that there are many ways how technologies can improve your workflow. How significant the impact of increasing data volume will be? For example if you will be able to trick not one mouse, but hundreds mice with the same amount of manual labor. Or, for example, if you increase the frequency of observations, will it move the needle?
V: It will depend on the specific study. And it’s complicated, because the rules by which we judge research are made by people and these are just agreements that were made. For example there are some spoken and unspoken agreements about how many animals should be in the study that you would consider it valid. If you use two animals, this is considered to be invalid. But it’s hard to say what is there perfect number. Normally the real number of animals involved is between what would be ideal and what would be realistic to do, and how many animals you are allowed to take by the ethical committee.
Because to do experiments on animals one needs to get approval from Animal Ethics Committee and you need to explain why you need so many mice for experiments and you cannot use as many as you want. A hundred mice, for example, will not be allowed.
P: Yes that’s so interesting. That’s very different from the way businesses work because there is a basic kind of assumption or maybe belief that the more data you get the better, and you do as much as you can in order to get it. Probably we see this kind of situation when tech is not the only one factor that limits data collection.
V: There’s one more thing about the number of animals in experiments. It’s a statistical problem that concerns repetitive measurements. So the more often you measure something the higher the chances that you will finally see some significant difference which is not there like you increase the possibility of false positive result.
P: So science and business have different views on data collection, sometimes. Science and business merge more and more and still they’re quite different.
Could you mention maybe two or three challenges from your field or from associated fields where demand for tech is clear?
V: So I’m sure there was or could be a better solution to do what I did and maybe do it more efficiently. Also it could be useful to iterate quickly and to compare different ways of analyzing the same data to see which one makes more sense. In my case I couldn’t really apply multiple algorithms and approaches because this was taking too much time. Therefore I had to decide what’s the optimum from what I see and act on that, without much trial and error cycles.
Analysing calcium imaging is something very important and it’s a growing field right now. It’s not only used to study diseases but also to study normal functioning of the brain.
Another theme in neurobiology is to analyze behavioral data. So if we give animals tasks and they do something we can also record that on a camera. And then the question is if the human has to look through the whole video and manually count all the different things that a mouse does or there is software that can analyze it.
P: It’s really amazing that you’ve mentioned the idea of analysing animals behaviour. It is something entrepreneurs also think about. For example, there is a startup that apply computer vision to analyse behaviour of fish at fish farms.
V: I’m thinking about the laboratory settings because that is what I’m used to, but there are also people studying behavior of animals in the wild and tech will be helpful for them. For example you can put a tracker on the animal, and people of course already do that. But I think they have very limited outcomes on that so I guess this can be improved. For example, if we talk about studying monkeys, now it’s about people observing them. But if you could put some trackers maybe on the limbs you can get data without a monkey being stressed from seeing a human being.
P: It will also free-up some time for scientists, who will be able to do more value-added things.
Part VI. Repurposing tech within the lab
P: What could be the ways of using the same technique, the same routines for other purposes. I can give you example. Imagine someone who is an astrophysicist and works with data from space which is very very noisy. Routines and approaches form astrophysics can be applied to data from financial markets, that are also very noisy.
I’m not saying about purely business or finance application in particular, but maybe you can repurpose your approaches to other research fields or other tasks within the same Alzheimer’s theme.
V: If we talk about thе method I’m using it can be very well applied in all the different topics of neurobiology. Whatever concerns activity of neurons and what whenever you suspect that the neurons are somehow altered because of the disease then you can use this method. You can also use it to identify what are the brain areas or specific types of neurons that are responsible for certain behavior.
People used this method to study better functioning of the brain for example in the visual cortex.
One more application is to explore fluctuations in different kinds of cells. In many cells there are some fluctuations happening. For example, there are certain fluctuations that are associated with circadian rhythm. So some proteins for example are higher or are more abundant during the day and less abundant during the night and otherwise There are hormones that are released at different times of the day and then they are also influencing the cells of other tissues differently.
So theoretically if we have an indicator of a change that we want to see and we use microscopy then we can observe these changes in vivo and see what it is like and describe it better and then maybe understand how you can manipulate this change.
P: How it could be done technically, via a blood test or a tissue sample?
V: If it has to be in vivo then it has to be within a body. If you take a blood sample or a tissue samples and it’s out of the body already.
If I’m being very futuristic… we make a very small camera that we can maybe swallow or we implant in a specific part of a body, basically we will need to do the same kind of step I’ve described with recording videos. We need to deliver an indicator to cells that we want to study and then we need to deliver a recorder of the signal.
It’s just too invasive because there are parts of the body where you cannot really get to non-invasive and if you use the light it’s always a problem because light gets dispersed when it goes through matter, maybe we find another way of transferring the signal. Imagine if the indicator is not light but is making a sound. Then we don’t need to get close to tissue and we can record how cells are ‘talking’.
P: I already feel my cells singing. Listening cells doesn’t sound too crazy. Just look at the progress in video recording, size of a camera, resolution etc.
Part VII. Science and entrepreneurs, research career
P: Let’s talk about interactions with computer scientists, those who, in many ways, make tech progress possible. Imagine a computer scientist, somebody who likes working with images, understands a lot about that, how he or she could start tackle problems you are working on?
V: What is important I think is not to act before there is a need for something. So ask people what they actually need before you build. For example, I would not make a software to analyze the ‘singing’ cells now. Even though it might be possible in the near future, it doesn’t make sense now, no one needs it right now. There are things that people need and where the help from computer scientists is very much appreciated.
P: As far as understood, you are excited about the way how science is organized, so you moved from research to manage and organise science. Tell about that, please.
V: Okay, so while working on my Ph.D. I realized that working in the lab and spending all my time there, I don’t really feel fulfilled and I feel that some of my skills are not employed. I figured out that there is a lot of other things that I actually enjoy and where I can bring value.
That’s why after finishing my PhD research, I decided to stay connected with science while also doing something using my organizational skills. Now I work as a trainee at the coordination office of the graduate program, that provides education to Ph.D. students. I also work as a consultant, helping people to navigate scientific careers.
I help undergraduate students to shape their careers in research, to identify which education they actually need, how to find the best programme, and how to apply there.
I think it’s very sad that many people think that if you decided to study a certain scientific field then your only option is a research career, being a scientist in a lab. And they ignore other options where they can also use their knowledge, do something interesting but at the same time different from working in a lab.
P: This is interesting that you mentioned these other options. One option that excites me in particular is entrepreneurship. I see an increasing amount of people who’ve moved from academia to startups, as founders, consultants, scientific advisors, etc.
What would be your advice for someone who is thinking to leave a research career, are there any indicators one can observe and understand that maybe she or he should change something?
V: I think they indicator is exactly your thoughts, the fact that you think that maybe this is not the best path for you.
P: Well, sometimes one can be just depressed. Maybe you’re tired. Are there anything else?
V: Okay so if it’s a persistent feeling that you’re doing something wrong then probably there is something wrong. If you are professionally doing something which is not very good for you, then probably you have some other areas in your life where you actually have a lot of fun and where you enjoy doing what you do.
You might feel that there are some things that bring you joy and there are things that don’t. Probably, the latter ones are not something you should be doing.
My recommendation for these situations would be not to leave everything behind immediately, not to quit your job. I think this is always just adding more stress on you if you feel that’s wrong and you’d immediately try to get rid of it.
I would try to find what actually you like doing. Maybe there is already something in your life which you enjoy, something you are doing in your free time, maybe volunteering experience. It doesn’t mean that each hobby has to become your profession. Try things before committing to them, ask people about what they are doing. I don’t recommend starting a new career before you actually tried it a bit. Many jobs look very attractive from the outside, but they are quite different from the inside.
If you want to become a scientist, before you quit your job and enroll to a university again, try to learn more about what is everyday life of a scientist, what actually science is about. Try to find a place where you can work with people on a small project as an assistant, or you can come and watch how they work. I think this trial is very important.
P: Your advice seems to have a flavor of ‘lean startup’ methodology, that is about trial and error cycles in some sense.
V: I think this just comes down to a completely scientific approach. I think also lean startup and agile techniques actually are very smart, because they are experiment-based. I think the experiment is something which can give you the best understanding about something.
P: Science and entrepreneurship share many things in common, so thank you so much for the talk and for highlighting several important issues.