 Welcome to the cutting edge of virtual reality for data science. The cutting edge is a new series in which I want to cover areas that I think show great promise, but are not quite there yet. Virtual reality has existed for decades. We've had head-mounted displays with motion tracking since the 1960s. However, despite that, there are surprisingly few actual applications of virtual reality. I believe the reason is a combination of expensive hardware, requirement for dedicated rooms to have it set up, and many competing systems resulting in all of them being below critical mass. However, that's suddenly changed with the Oculus Quest 2, now known as the Meta Quest 2. It's a standalone device, which has already sold in the order of 10 million units, meaning that it's readily available and VR is now here. So is it actually useful? One of the first things people tried was to make virtual reality meetings where you have a virtual venue and you can move around and look at things. However, the one thing you cannot see in virtual reality is other people's faces because they're wearing a virtual reality headset. People have also tried making virtual reality offices, allowing you to run any kind of software in a virtual environment, including all browser-based tools. This means that you can be at a fancy space station, fire up a web browser and work with your favorite tool, the string database. However, let's address the elephant in the room. The windows are still 2D. I don't really see the point of doing this. So what do I think it might be useful for? Well, looking at data when you're a data scientist. Let's start with the obvious, 3D data. If you're working with protein structures, 3D genome organization, anatomical models, or maybe 3D imaging data, you're dealing with data that are 3D by nature. In that case, there are obvious advantages of using a virtual reality environment for looking at something that is 3D in 3D. And for this reason, many tools already exist. BioVR and proteinVR are both tools that allow you to look at smaller and larger molecules in a 3D VR world, and UnityMool is a newer tool for doing the same thing with arguably more impressive graphics. There's also BabelVR, which has been designed to look at medical imaging data inside a VR setting. But what about high-dimensional data? Most of the data we deal with in data science is, after all, not 3D. A common approach is to do dimensionality reduction and do 2D visualization of the data. And it is, of course, trivial to go from 2D to instead 3D. The advantages are much less obvious, though. If you're already reducing the data from a thousand dimensions, is it really a big advantage to have three dimensions instead of two? Well, people are exploring these approaches, looking at, for example, single-cell expression data, having developed tools like SelexoVR that allow you to look at these data having 3D UMAPs instead of 2D UMAPs, and much more. Similarly, people in the network visualization field have developed tools like VRNetser that allow you to look at very large protein interaction networks and other networks in a virtual reality setting. Both of these tools are full-featured tools that allow you to visualize your own large datasets, query the data, and inspect individual data points. How useful they will be, the future will tell. So are we there yet? I think we still have some way to go. The problem is that there are very few native applications so far. The tools I've talked about today almost all require a Windows PC with a powerful graphics card to run. And let's face it, most of us do not have that sitting on our desks at work. In the future, I believe tools will be designed to run on cheap VR hardware and be based on open-web standards, meaning that they can run on any current or future VR headset. Protein VR is a good example of that, but I believe there will be much more to come in this direction. If you want to see more about what one can do with virtual reality and visualizational data, take a look at this presentation next. Thanks for your attention.