In this fascinating episode of The Omni Show, we invite cognitive neuroscientist Dr. Matthew Kmiecik to discuss his career, his workflow, and how he uses OmniGraffle to make data more accessible and engaging. We dive into Dr. Kmiecik's background, his evolution from neuroscience to data science at 23andMe, and the impact of his published academic research.
Dr. Kmiecik highlights the importance of effective data visualization and shares his unique approach of combining data science using the R programming language with OmniGraffle. He reveals how he uses R to create initial visualizations, then imports them into OmniGraffle to add finishing touches. We also chat about saving data visualizations as SVG files, allowing for increased flexibility and control when editing in OmniGraffle.
Some other people, places, and things mentioned in this episode:
- R Programming Language
- GGPlot 2
- Dr. Matthew's LinkedIn page
Andrew J. Mason: You are listening to the Omni Show where we connect with the amazing community surrounding the Omni Group's award-winning products. My name's Andrew J. Mason, and today we learn how Dr. Matthew Kmiecik uses OmniGraffle. Well, welcome everybody to this episode of the Omni Show. My name's Andrew J. Mason, and today we are beyond thrilled to be able to have Dr. Matthew Kmiecik with us. He's a cognitive neuroscientist doing his second postdoc at 23andMe, and he's researching various markers of Parkinson's disease and using OmniGraffle every day in his work. Dr. Matthew, we are thrilled to have you with us. Thanks for joining us.
Dr. Matthew Kmiecik: Yeah, thanks for having me.
Andrew J. Mason: Well, Dr. Matthew, I'd love for you to tell us a little bit about how you ended up where you are. This is such a really specific career path, cognitive neuroscientist at 23andMe, so maybe breakdown how you got to 23andMe and then also how you ended up in your particular field of study, Parkinson's disease. That's actually fascinating.
Dr. Matthew Kmiecik: Sure, yeah. I love to talk about this. I speak sometimes to undergraduate students who were in my position several years ago and they're trying to make a decision in their career path and sometimes it helps hearing where I came from. So I went to college, a university called Loyola University of Chicago to become a medical doctor. I thought I was going to be a pediatrician since I was in about in fifth grade. But while I was at doing my undergraduate education, I discovered research. I was told that starting in a research lab, being involved in the process of running experiments and analyzing data makes you a really good candidate for medical school. So as soon as I got into Loyola, I tried to find any opportunity I could into research labs. And it wasn't until my junior year really where I really found a cool cognitive neuroscience lab that was studying human analogical reasoning and recording brainwaves at the same time, which I thought was super cool. And so I was doing my best to take all the classes and study for the tests that were needed to go to medical school but I found myself just really wanting to play with data, run experiments, come up with new ideas on how to test the human mind, how to study it. And I was just finding myself just staying up really late, working on data analysis projects and analyzing data to the point where I was like, man, I should probably start studying. And then I started to make the realization after just wanting to play around with data all the time that, hey, maybe I should pursue this as a career. So I had to make this sort of change into maybe medicine isn't the right path for me, maybe it's studying the human brain. And so I made a decision there to stop pursuing medicine and going into cognitive neuroscience programs. And so I then started applying to PhD programs that do similar research that I was interested in. And then, yeah, it took me down this crazy career path where I've actually been really fortunate to study a variety of different things. I've studied traumatic brain injuries, I've studied how we derive meaning from reading of sentences, also recording brainwaves at the same time. After that I studied chronic pelvic pain and how menstrual pain and women can convey elevated risk for chronic pelvic pain and specifically sensory processing. So how we process visual information, auditory information, our visceral information, so those things from our bladder and how all that kind of pain processing combines into one to kind of conveys like a pain response or perhaps the chronic pain response later in life. And then now I decided to make another kind of change in my research trajectory and study genetics, which I have no background in, which has been great at 23andMe to get this kind of experience. And as well as Parkinson's disease, which is the number one growing neurodegenerative disease in the world. I've taken a crazy career path, but throughout, it's been just fantastic.
Andrew J. Mason: I'm really excited to dive into some of the research and your findings in 23andMe or in past postdoc studies. I'd also like to know if you have any recollection at all of first interactions with the Omni Group or OmniGraffle, do you recall how it entered your world or showed up in your space?
Dr. Matthew Kmiecik: Yeah, so I went to graduate school at the University of Texas at Dallas. And when I was studying there for the cognitive neuroscience program, the lab that I started working in, they were using OmniGraffle to create their stimuli for their experiments. And so they were interested in how humans do analogical reasoning, so finding the similarities between differences of things. And when they were making the stimuli for that project, they were using all of it in OmniGraffle, which was this new program for me. I was not even a Mac user when I started graduate school. They had these big IMAX in the lab and they had OmniGraffle on there, and that's what they were using to create the visual stimuli for the project. So usually what we'd do is we use OmniGraffle to create the stimulus, and then we'd save it out as either like a PNG or JPEG file, and then we would fire that up into another program where human participants, they would come into the lab and look at these images and provide responses for us. And so OmniGraffle that was my first sort of introduction into it, we were using it for one project. And then afterwards, after we're done with that project, I kept on gravitating back towards it because I was like, oh, I could just easily do this in OmniGraffle really quick. And that's basically how I got introduced to the software.
Andrew J. Mason: Wow. Okay. So OmniGraffle used to create the stimuli in clinical studies. I've heard a lot of use cases, that's the first use case I've heard of that. Fantastic. What other spaces or use cases have you seen OmniGraffle used for?
Dr. Matthew Kmiecik: Yeah, I don't know many other people outside of my research circle that use OmniGraffle. So I'm probably just really specific about the use cases here but what I've noticed being used for is mainly for the development of scientific posters. So academic posters, when you go to a conference, you want to present your data to the public or those that go to the conference. And creating a poster in OmniGraffle tool is great because you get to specify the dimensions of the paper that you want to print it on. So usually we print it on like 42 inches by 54 inches, these big kinds of poster boards. And so that's one use case for it, which is academic posters. Another one is for figures for manuscripts. So these could be either visual depictions of the protocol. So what a participant will see when they come into the lab, when they're participating in the cognitive neuroscience experiments. I've also seen it used for depictions of clinical trial progress. Let's say you have a group of eligible participants for a clinical trial, you enroll this many, and then from there, this many are eligible, and then they go into various arms. Usually that's called a consort diagram. I've seen OmniGraffle used for that. As for me, I typically use OmniGraffle for the development of manuscript figures. So when we're writing up our paper for publication from our science experiments, we'll go ahead and develop those images and those figures for data visualization in OmniGraffle because of its really great export tools and just manipulation within the program itself for saving out figures. And then almost all of my scientific publications have figures that have been finalized in OmniGraffle and then eventually published.
Andrew J. Mason: That's really cool, Dr. Matthew, thank you for sharing that. I would love for you to share some of... and it's not necessarily OmniGraffle related per se, but what are you finding in your research in Parkinson's with 23andMe? What are some of the things that have showed up for you that you deem, "Hey, this is interesting, we should pay attention to this?"
Dr. Matthew Kmiecik: Yeah, unfortunately, I don't think I can talk much about the 23andMe results yet because they're unpublished but if I could talk about perhaps some of my previous published work, I could talk about that. How does that sound?
Andrew J. Mason: Yes, that would be amazing.
Dr. Matthew Kmiecik: Okay. Yeah, sure. So in my previous postdoc, we were really interested in visceral pain including menstrual pain and bladder pain and how that conveys greater risk for development of chronic pelvic pain, increased pelvic pain long term. So we're talking about maybe one to five years after their study visit. What we were seeing there, and this is from a lab, Dr. Kevin Hellman and Dr. Frank Tu's lab in Evanston Hospital in Illinois, they were basically seeing that if you have an increased sensitivity to light, increased sensitivity to sound, you have elevated menstrual pain, you have elevated bladder pain, you have elevated sensitivity to touch around your body. All of these sorts of predict an increase in chronic pelvic pain later. So as participants would rate their pelvic pain a year later, two years later, up to five years later, we would see that initial sort of sensitivity at their baseline visit was predictive of their pain even four years later. So it's sort of suggests that there's like this overall central sensitivity measurement or metric that we could use to really predict somebody's future pain. And this was actually more predictive than their pelvic pain at their baseline, which I found really interesting that something else was predicting their pain versus just asking them, how much pain are you in right now? Because usually how much pain you're in right now is going to be predictive of your future pain. That's sort of what we were finding in my previous postdoc, which is all those results are just published. I just recently published a paper in April this year in the journal called Pain, which people can read about that if they're interested.
Andrew J. Mason: What advice might you have for somebody who possibly recognizes you as a leader in the space of data visualization and they want to get started with data visualization in some way, but they're just not sure what that first great step might be? Where would you direct somebody who had that kind of a desire?
Dr. Matthew Kmiecik: Yeah, that's a good question. I think that what really got me interested in data visualization was reading the works of Edward Tufte. I read some of his books on data visualization and just really fell in love with the idea of data visualization needs to be really clear, needs to be concise, need to convey a lot of information in a small space. Just make it really easy for the reader to understand what you're trying to say. And then actually being cognizant and aware of developing good figures, good data visualization for your audience. I think that's enough right there just to get you really thinking about, okay, well, I need to create this figure and how's the best way to do it? And by reading those works and maybe playing around in OmniGraffle a little bit, you get a sense of, okay, is this clear or is this not clear? Getting feedback from other people is also really important. So showing this figure to somebody and saying, "Hey, what do you think I'm trying to convey here?" And taking that feedback seriously, because I think, at least for my use case, whenever I'm developing data visualizations for publication, when somebody's reading your paper, the first thing they're going to go to is the figures, right? They're going to try to understand things in a visual way rather than reading all this scientific text and jargon that is being used. And so if you can create very clear data visualizations and infographics, it makes things a lot easier on the audience. And so I guess the number one tip I have is to be empathetic for your audience. And if someone's not understanding your figure right away, it's probably a good idea to make some changes to it so it's constantly like a work of art. I see a figure kind of a slab of marble. You're just constantly molding it and molding it and molding it until the point where you get it to be really crystal clear to your audience. Another thing that I would probably have to suggest for beginners is... I guess how I sort of got started in it was I watched a few YouTube videos of people doing projects in OmniGraffle and showing the ropes, and that's how I got familiar with the software in the beginning. But I think the best piece of advice I would have for somebody is throw yourself into it, have a project in mind, and try to accomplish that. And most things are possible in OmniGraffle reading the forums, whenever I got stuck or I didn't know if something was possible, then I would just search the forums, OmniGraffle's website, and I would find a lot of cool tips from people that are trying to solve those problems. It's come in handy for me a lot, just making a lot of different diagrams. And then also, I've used it for mind mapping. When I would start to learn about a new topic, I would take a text box in OmniGraffle and put the concept there and then have arrows and diagrams going all different directions to different concepts that I'd be researching and just to get my thoughts on the paper and how they sort of connect with each other. Sort of like a crazy detective having strings, connecting pictures all over the place but it's nice because it's digital and you can move them around and the arrows kind of connect the boxes together and they're fixed in that way.
Andrew J. Mason: As you were sharing that, this is something that just showed up. I'm curious about what indicators you use to decide what form of visualization data should take. So when you see information or a pattern emerge or some way you know okay, this needs to be presented visually, is it more of an art versus science like you mentioned? Or are there key things that you look for to say, hey, this is a bar chart versus this is a line graph or whatever, but there's some way that you know how data should show up. Can you speak to what you used to make that decision?
Dr. Matthew Kmiecik: Yeah, that's a really good question. I think a lot of it is from experience and just thinking about what kind of... what's the best way to display the data. I do remember from graduate school, one of my first classes I ever took was in statistics taught by Dr. Robert Ackerman at the University of Texas at Dallas. And I had this moment where everything I thought I knew about making figures and scientific publications was wrong because he basically was like, okay, so here are some data you want to plot, let's say the mean accuracies from some kind of task from a group of individuals. And so a lot of us would immediately go to the bar chart. You see these bars, you see that one group has more accuracy than the other one on average, and then you put some sort of error bars on there, whether it's a 95% competence interval or whether it's standard error, the mean, or something like that. And he went to explain that the bar chart, just by having physical bars on the chart that sort of signifies that that's a very fixed value, it's a plateau, it's not moving, it's very fixed. Whereas if you were to replace those bars with a point, it allows the reader to infer that perhaps that value is just not just a fixed value, but we just took a sample of let's say 20 individuals that gave their response, and this is what we estimated, but those that studied the central limit theorem know that that estimate is going to change with the population, and that's why we have an error to it. And so subtle changes like that made me think that oh, what we are showing with our data is you're suggesting subtleties to the reader. And so having that in mind while you create your figures, I think is really important.
Andrew J. Mason: And I think our heart here really is to help people be productive by learning from us through the positive aspects and maybe some of the aspects where it's like, if I had this to do over again, I would do it differently. Is there anything that you can see from your career track thus far that maybe it's not a failure, but it's something that I would just skip that slice and you might be better served just not going in that direction or doing that thing or thinking in that way. Is there anything that really fits that criteria for you that maybe would be instructional for others?
Dr. Matthew Kmiecik: Something that really accelerated my workflow, I don't think I've talked about this yet, but my workflow is basically I will use a statistical programming language called R to analyze the data and create visualizations. And from there, what I would do is I would take that information and put it into OmniGraffle where I would do small tweaks here and there to make the figure really just pop and really convey a lot more information to the audience than it would just coming straight out of R. And the package I used to do that in R is called ggplot2. And so ggplot2, what I used to do was I would take that image and I would save it out as a JPEG or a PNG, and then I'd pop that into OmniGraffle where I would make some adjustments to it from there. But what really propelled me forward in terms of making my work a little bit faster, a little bit more amenable to changes was to not worry about getting the figure a hundred percent correct in ggplot, which is the package in R that I was talking about, but getting that figure in a state of like, okay, this is really almost at its final point, but there's some changes that I want to make in OmniGraffle to it to really just convey some more information and saving that out as an SVG file and then putting that into OmniGraffle allows you to have a lot more flexibility with the program. You can make a lot more minute changes in there. And if you were to, I guess, make some additional changes to the overall figure, you do have to go back into R and make those changes. But in general, I have acquired a lot of feedback at that time from other people. I've gone through several iterations within the software program, and so by the time it gets the OmniGraffle, it's in a pretty good state. And so from OmniGraffle, I really use it to make the finishing touches and to add some additional things that it's a little bit harder to do in R, it's a lot more time-consuming than I would just throw it into OmniGraffle to make that job a little bit more efficient. I guess that was kind of the light bulb moment for me. It's like, okay, I should be doing this in a different way, and that really made me more efficient.
Andrew J. Mason: So that file saving from SVG allows you to stay within the OmniGraffle environment versus having saved the file as a rasterized something, and then it's just frozen in place.
Dr. Matthew Kmiecik: When you put it into OmniGraffle, you are able to edit every individual point. And I know this is highly niche to people who use R as their programming language and they're saving things out. But I think this, it also works for other things too that I found that are saved in SVG format. If you put them into OmniGraffle, you can edit every little minute detail within that figure, and you can really just make it your own. And so that's what I really love about it, because you have just so much control over what you're doing, and then when you go to save it out, you can save it out into different file formats that you want. And for me, I guess for academic publishing, it's really important to have those figures in a very crystal clear high DPI image. So when they go into print, they print really nicely and they're not blurry. And so I guess that's another part of my workflow that I really appreciate that you can control those aspects. Anything that I have to do with figures or images, whether it's for just something really easy, maybe cropping an image instead of going into other programs, I'll just use OmniGraffle because there's just so much that you can do in there.
Andrew J. Mason: I love it. Thank you so much for your time, Dr. Matthew, and just spending it just sharing with us what you're able to accomplish using the software, but also just kind of giving us some tips about how to best visualize data. I think that's really cool. How can folks connect with you and what you're up to and stay within your orbit if they're interested in finding out more about what you're up to?
Dr. Matthew Kmiecik: Yeah, I have a website. People can go on there to see all my publications and all my posters and some of my blog posts that I'm writing. So if you can go to mattkmiecik.com, that's where you can find me. I'm also on LinkedIn, and that's a platform that I use quite often, so you can find me there as well as on Twitter as well. So those are my main platforms that I'm using but my website will have links to my publications in which I've used OmniGraffle to create the figures for, and some of my blog posts actually feature figures that I've made and polished up in OmniGraffle.
Andrew J. Mason: Perfect. Thank you so much, Dr. Matthew.
Dr. Matthew Kmiecik: Thank you so much for having me. Thank you.
Andrew J. Mason: Hey, and thank all of you for listening today too. You can drop us a line on Twitter at the Omni Show. You can also find out everything that's happening with the Omni Group at omnigroup.com/blog.