View transcript
Thank you very much. Peter stole my first line there, which was to explain why I put so much detail on the slide. It was to point out that I work in a biological area, but I'm located in a games technology department, which I think is quite an interesting mix. So, medical visualization. It does work if I lean a little. We had a technical hitch to do with my laptop not being as good as the others. You can use visualization for information. You can use visualization, it really does need this, for planning. And I mean sort of medical planning. And you can use visualization, that one worked, for experimentation. It's the latter I want to focus on. You came here wanting to know about VR and various bits of technology. See, every time. This isn't too happy a slide. So, obviously, it's UK based. I didn't get a lot of time to put this talk together. So, one in two people in the UK, and actually Denmark and most northern European countries are not particularly different, will get cancer in their lifetime. That's not a happy statistic. No? There we are. So, cancer. So, the question arises, and we can give a tiny bit of cell biology here, just so the talk makes sense. When normal cells start looking a little bit weird. And the reason they look strange is because their behavior has changed. It's gone into an aberrant state. A crazy state. Where the cells start dividing more quickly than they should. They grow bigger than they should, and they don't die when they should, or some mix of those. or at least part of what's causing it, is what's called the cell wiring diagram. So inside the cell, you have the cell wall around the outside, the nucleus in the middle which controls what's going on, and then right around the edges are the receptors. These receptors are little listening beacons that listen to what the body is telling the cell to do. Those cells then, the signals then propagate down through this network to control the nucleus. What happens in cancer is a little bit like if you have a faulty wiring diagram in your house. You turn the light switch on in the kitchen and it switches the lights off in the bedroom and turns the microwave on. That's what cancers do. And it's very difficult to understand partly, at least these pathways, sorry, that control whether the cell lives, dies, and actually divides. But it's actually more like the wiring diagram in this building and any one part of it can be very, very problematic. And the key issue actually is that a large part of that network you can change and it won't change the cell function. You change one part of it a tiny bit and you get a huge change. So it's a really difficult system to work with. Cancer drugs work in what's called Petri dishes, soup of cells, cancer cells. One of the good things about cancer cells is they're good for experimentation because they live forever. And they're quite hard to kill off. So the very thing that makes them bad makes them good to work on. And what you do in cancer drug discovery is you take your Petri dish and you put some drugs in, you mix the drugs through, and then you measure what those drugs do to the cells and try and understand that biologically. It costs about a billion euros and takes about 12 years to get a drug from the lab to the market. And it's a key problem. You're looking at maybe 10,000 plus trial compounds to get to a single drug at the end. That's a huge waste. And it is in effect a big search problem. How do you find out good ideas in order to get that one drug? So part of it is understanding the system. So you can use visualization for information. So this is a video, it's not from our work, it's from another group, looking at cell division. Now I hope this makes more sense than the explanation I tried to give you. Because it turns it into a picture, so it's clear what's going on. It then talks about how it was done, and I think I've stopped the video there. This one here is to try and highlight cell signaling. So you can see the signals are happening and then the nucleus splits, the cell grows into two, and then it divides. And that's what's going on within cancer, and that happens at a rapid rate. So you can use this to try and explain to people what's going on. You might use it for an educational perspective, for patients, you might use it when you're teaching lower year university students and so on. You can use the same kind of interactive visualizations for all rather sort of... I didn't mean for the sound to come on, but it's okay. It's not the most, unless anybody's really into chemistry, biochemistry, no, we don't need the sound. Right, that's a much more technical video which you can use to explain how two things interact. So you can use lots and lots of interactive visualization for information and that's fine, but that doesn't really float my boat. This is a bit more interesting looking at planning. I know from speaking to some of my surgical colleagues that they spend a long time talking about a particular operation. Sometimes they'll do it using scans. For cranial surgery, for example, or jaw surgery, they will take a scan, they will print a plastic model of that scan of their head using 3D printers, and they will, and they will go at it with drills because it's really important that people know where to stand for the right bits, because obviously you only get one go at this, right? So they'll even print a couple of models and they'll line up and they'll stand over it and they'll practice. Once it gets past that stage, it's kind of down to being cubits, they're nuts and bolts and sticking things together. But, so they practice with plastic printers, but you could also equally use the HoloLens, I hope. We've got sound on this video as well. My name is Annie. That's not, I'm not Annie. The point here, if I could pause this, this would be good, is that you could, you can walk around this and the surgeons can stand over the work and they can prepare, not near the edge, they can prepare the surgery, they can practice, they can talk about it, they can look at a particular condition and go, if we go in this way, that's okay, I'll just, it's nearly finished. So you could use that for surgical planning, which is great. And I think there's a real opportunity for AR in this. Groups of surgeons talking around a particular operation. I'm not particularly interested in that either. What I'm more interested in is in this. Using virtual environments, and this will move into VR as I progress, for experimentation. So I spoke about the cancer cell, that's exactly the same, that's exactly the same diagram as I showed you before just, it just zoomed in a bit. You've got these pathways that control cell survival. Cell proliferation, so when the cells reproduce. We can make a mathematical model of that because we can measure what's going on inside the cell. But model analysis is quite hard because that's just the picture of the network. That's like a train network without any of the time table. You need the time in there. time in there as well. And when this is 60 plus differential equations, it becomes quite hard to understand, unless you've got one or two degrees in maths. But it's actually worse than the maths. This is the same network, just drawn in a different way. You get what's called crosstalk. So things happening on one pathway affect things happening on another pathway. These drugs, they target particular nodes here. So if you were to try and block activity down one pathway, that can cause an increase in activity down the other, because it will feed back through the crosstalk. You also get feedback loops. So sometimes you can downregulate an activity through a drug, and it will cause a sort of congestion lower down that then feeds back and maybe accelerates something further up the link. It's also not a static network. This is probably the worst thing. So combination therapies, you may have heard in the press, this is one of the latest bit of thinking in how to treat cancer, combination therapies. What you can do is add in two different drugs. Now, these two drugs, I don't need to get into detail of this, but they work in different ways. This one blocks the cell from growing. This one damages the cell DNA. So they're working together. But it turns out that if you just give, let's say, the red drug, the growth inhibitor, you don't get a very good response. Not many cells die. Apoptosis is just a fancy biology name for death. So if you give two drugs together, you get some response. If you give the blue drug first, you get some response. But if you give the red drug first, before the blue, you get a very, very strong response. So what you can do is, by changing the order and the dosage, increase or decrease the effect that the cancer drug has. So great. All you have to do for cancer is give the red drug before the blue drug. Well, sadly not, because if you do that, you're going to get a very, very strong response. So if you change from one cancer to another, or one patient to another, this also changes. So it becomes an individual problem as well. So it's a very, very difficult problem. You've got crosstalk. You've got dynamic rewiring. You've got complexity. You've got feedback loops. I work in a games department. What can we do to help? Well, we can start to try and unravel this a little bit. And so what we've built, and I don't normally show this, but I am doing it this time, is a signaling visualization toolkit. All right? We want to be able to draw this in game space. And it's a mix of arts, media, and computer games, which is where I'm from, but I'm also sort of associated with the science and engineering school and the school of medicine as well. So it's only by working together can you actually get this stuff done. What we've done is turn that rather dry railway network into some interactive movies which show the response of the network to the drug. And here we're going to add in a drug. This is a growth inhibitor drug. We can choose the dosage of the drug. We can add it in at whatever time. And once I press play, what this shows is that the network responds. It cools down because that's what the drug is doing. It's a growth inhibitor. So you see a massive drop in activity. But think of it as a traffic congestion. close off a roundabout or roadworks, if you close off a road, you get a bottleneck of traffic somewhere else, and that can cause problems in itself. So what this does is show how one simple intervention can have a whole range of different effects. Some stuff increases, some stuff decreases, some stuff doesn't change. And this is calibrated on biological data, so we can use this to dial in dosage without doing expensive real experiments. We can also add in other drugs. So we're going to add in, this is to introduce a mutation in the model. So cancer is often caused by mutations within the biology. We're going to introduce from that massive list one number, and we're going to change that number from a 40 when the model starts to a 20. That's the only change. We run it, we compare, side by side, see what happens, and what you get is a network that barely responds. Now I could have changed 30 other numbers and it would have made almost no difference. That one number, and it changes everything. And we can only begin to understand this by doing clever searches with a computer, but it's too big for AI. We need to use real AI, but not computing people. Computing people don't know about this stuff. No offense, Paul, in the audience who helped program this, but your knowledge of cancer is not what it should be. We need to take, we need to put the model in the hands of the biologist, and that's what this sort of technology that we've been talking about all day can do. It takes problem domains into the hands of people who know about them and can work with them and solve the problems. This, I think, just shows that you can rotate the network around you, but you'll all know this because of your background. But when I show this to some biologists, they're like, oh, I didn't realize you could do that as well. So, you can turn it around. You can zoom in and out. You can rotate the network just to get a better look at things. Some things get the wow factor from different audiences. But that's the point, right? So, where's this going? The big, big problem with cancer is not what I've shown you, actually. It's that cancer's not one cell. A cancer tumor's maybe half a million cells, a million cells or more. The real problem is they're all different, and every single one of them has got one of those signaling cells in its blood. And even if they are a cancer tumor cell, then that two million at least,� dank you got there. Thank you. technology to build a virtual tumor with a half a million cells, just like the one I showed you, so we can understand what the consequences are of individual cells being different. We can't explore that easily, and this is where we're going to recourse to VR. We want to get into that tumor and fly around and look at it and see how the drugs are sort of pervading from the outer tumor down to the middle to see how the individual cells are going to respond. Cloud-based computing. The last thing I wanted to mention was some work that is in very early stages. Early stages in terms of thinking, not in terms of time, but because it's too hard. So NARCS, which is an interesting name, it stands for Narrating Complex Systems. Book to come out soon, apparently. Complex systems struggle from a problem. Every time we write about them in a paper, it becomes really boring because you have to write about this really interesting system in a linear narrative. When you try and describe a complex system, if you think about a flock of birds, this is a classic complex system. You watch the flock of birds and it swoops and it goes in and out and it swirls through the air and it's a remarkable thing. How does biology try and understand that? By chopping a bird up and seeing it. By seeing how it works. You're losing the overall complexity because we tend to strip out in our narratives in order to be precise. So somehow we've got to find a way of making use of non-linear storytelling because all of these systems are non-linear to try and understand all of these complex systems. And so we've got another project starting soon which is going to try and explore this with, again, it's focused in cancer, but it's trying to find ways of telling the story. Stories as to what's happening with patients, but building on those cancer simulations. And again, we're expecting VR to play a big role in that, to try and help patients understand what's going on. There's quite a lot of work in patient decision making. It's very difficult when you've got cancer to make the right decision for treatment, partly because you can never know it's the right decision. But when people are, when they have the information packaged up as facts and figures versus narrative and patient experiences, patients are much happier, they're more content with the decisions they've made based on the narratives than they are based on the facts and figures. The decisions may well be the same, but it's that they've understood the problem space better so they're more comfortable. And again, I think VR has got a role here. I would want to acknowledge my many colleagues from, who were involved in some of that presentation around the cancer work. And I'm happy to take any questions on what I realised was probably a slightly left field talk. You probably weren't expecting a bit of cancer biology when you came in this afternoon. So happy to take questions. I knew that would be good. Thank you. Thank you, Jim. That was excellent. So any questions for Jim? There's one at the front. There's one at the front from Joe. Thanks, Jim. You know what, this is brilliant. You know that figure that you mentioned in the slide where you said there's something like one billion pounds, 10,000, one drug? Oh, I'll have to go all the way over there. The problem with using many, many quickfire slides is that when somebody asks you a question on one of your many, many quickfire slides, you have a lot to go back through. This one. Yes. That wasn't bad. 10,000, one drug. Yeah. Yeah. One drug. Yeah. How long does that take? That takes about 12 years. 12 years, yeah. So with the new simulation, is there a cut down in time or is that still the timeframe that you'd be working to? The simulations are still operating in this time window, so it's not had the opportunity to propagate through yet. This is the time consuming bit, so the six to 12. This is where you have to go through the clinical trials. With the clinical trials being complemented by the computer simulation or going into VR simulation, do you see that time being reduced? No. I think the time that gets reduced is this bit. I don't think that there's a, for the type of approaches I've shown you now, it doesn't have a role here, but it does here. Yeah. Yeah. Yeah. And you can't really predict what the best 250 compounds might be from the 10,000. And so when you start to think about this bit, simulation can't help with this bit, because it's not accurate enough. It doesn't have the information around real patients. It's based on cell lines. So it's a, it's a kind of a rough guess. So it still has to be worked out with. Absolutely. Yeah. Real trials. And actually, one of the challenges that there's been a few talk, a few mentions of the role of this in mental health and dealing with ADHD, et cetera. One of the challenges there is that there isn't not yet a single study, that I'm aware of at least, that actually goes through a proper clinical trial with a randomized test without any bias in it at all, which would demonstrate clinically that games and VR might help with mental health. There's lots of anecdotal evidence. There's semi-structured trials that help. But you cannot reduce this. Otherwise, you're into unsafe medical practice. There's a real switch here. This is working on mostly lab. And I'm sorry to say, sometimes it's working with mice and other animal systems. That's not what we do. But for some, it's a necessary step. And again, the models can help there. But as soon as you get into the real patient, that's then medicine. That's certainly where I would draw the line anyway. But it is an interesting question. Any more? Okay. Well, thank Jim once again for stepping in like that because we had a 15-minute slot. What a great way to fill it. Thank you, Jim. Thank you for your attention.
