Bhramar Mukherjee: A data-driven fairytale
Bhramar Mukherjee, professor of Biostatistics at the University of Michigan School of Public Health, explains how biostatistics helps make sense of big data for disease prevention and treatment and how she encourages her students to stay patient, optimistic, and attuned to their own inner voices in their pursuits. You don't need to be a math person to enjoy her journey.
Mukherjee shares her professional journey, from her math-centered upbringing in a family of liberal arts scholars to her roles at Purdue University and the University of Florida, eventually leading her to her academic home at the University of Michigan School of Public Health. She discusses her involvement in launching an undergraduate summer program on big data and her work in modeling COVID-19's trajectory in India.
Host: Mathematics and Human Health may seem an unlikely partnership at first glance. Math tends to be logical and based on absolutes, while the pursuit of human health is often a messier affair. But there is a sweet spot where statistics and public health meet, known as biostatistics, and in this age of information and big data, a partnership thrives. Biostatisticians create methods for combing through huge sets of data. The human genome is about one good example. A single genome is comprised of nearly three and a half billion base pairs. With statistical modeling tools and techniques, biostatisticians help researchers find the gems hidden in these overwhelming sets of data, which leads to breakthroughs in disease prevention, medical interventions, and even helping to predict the next pandemic. By the way, if you're not fond of math and statistics, don't be scared off, though there will be mathematical talk in today's episode. You don't need to be a math person to enjoy the journey.
Bhramar Mukherjee: There have been times when people said that, well, quantum computing is going to oust statistics. Blockchain computing is going to oust statistics. Machine learning is going to take over statistics. AI is going to take over statistics. Operations research is going to take over statistics. But we are here and we are going to stay here because we bring something very unique to society about studying natural phenomenon with a framework of design and analysis, how to properly design studies. Big data cannot lead to credible knowledge and actionable knowledge unless you understand why an inferential result is appearing.
BM: Sitting in a school of public health, my job is not just to predict, but to prevent. And in order to prevent, I do not just need a mindless, big black box giving me prediction. I want to understand causal mechanisms. I want to understand the pathways. I do think there are challenges, but the opportunities are far, far outnumber the challenges. And even at age 50, I'm looking at the next 20 years of my career of learning and really growing in a very volatile and very growing field of quantitative sciences.
Host: Hello and welcome to Population Healthy, a podcast from the University of Michigan School of Public Health. Join us as we dig into important health topics, stuff that affects the health of all of us at a population level. From the microscopic to the macroeconomic, the social to the environmental, from cities to neighborhoods, states to countries and around the world. Bhramar Mukherjee is the John D. Kalbfleisch's collegiate professor of biostatistics and chair of the biostatistics department at the University of Michigan School of Public Health. During her impressive career, this self-proclaimed math nerd has co-authored more than 370 publications in statistics, biostatistics, medicine and public health. She served in many capacities and settings and was recently named to the newly created position of Assistant Vice President for Research Data Services and Strategy for the University of Michigan. Mukherjee sat down for an interview about her career and some of the experiences that helped shape her passion for using the rigors of mathematics and statistics to improve the health of the world.
BM: I grew up in a family of liberal arts people. My father is a very celebrated actor and a public intellectual in India, my sister studies American literature, my mother studies Bengali literature, our entire family is students and scholars of classics and play and poetry. So I was the odd person out. I grew up as a math nerd and I'd sit at a dinner table and hear all of this confusing debates about whether a poem is a good work of art, whether a movie qualifies as classic and people will eternally argue without any resolution. So this bothered me at an intrinsic level that there is no objective one answer. I looked for singular answers and solutions.
BM: So I started to do math because math, regardless of what I do and who the instructor is, if you get the answer right, you have to give the points. I studied mathematics and my family was not quite happy about it. So I had to hide all my math and physics books because my father wanted me to study Rousseau, Tolstoy, Shakespeare and we had our reading list and it was very hard to manage because there was no Archimedes or Newton or Gauss in that reading list. But I do think this liberal arts education gave me a very wonderful background to explain my work to others and also think about mathematics not just as an very abstract notion but try to think about statistics more as a derivative of philosophy, trying to explain natural phenomena that are happening around us rather than a mere derivative of mathematics which is much more structured. So I think this passion for mathematics and objectivity combined with humanities and social sciences led me finally to public health and biostatistics.
Host: Mukherjee moved to the US from her home country of India to begin graduate studies in mathematical statistics and probability at Purdue University.
BM: Purdue at the time was a very theoretical department. There were very few applied courses. We were still housed with the mathematics department and computer science was in the vicinity so it was much more of a computational and mathematical environment than translating your work to society. So in that environment I looked for opportunities to see the connection of what I was studying to real problems, actual problems that were going on. And at the time there are consulting opportunities in Purdue which I availed. We were working with community organizations who do not have resources to hire a data scientist or a consultant or an analyst and all the graduate students at Purdue statistics were helping campus partners and local organizations with pro bono statistical help.
BM: Statistics have this amazing ability to play in everybody's field because we have a common language of translating problems into models and into inference and quantify uncertainty. So I feel very privileged and honored that I had the opportunity to work on things beyond my dissertation and that's one thing that I have always advised to my own students that dissertation is only a part of you but try to get engaged in other problems which stimulate the other parts of you because the holistic education is extremely important.
Host: After earning her PhD at Purdue, Mukherjee began her teaching career at the University of Florida.
BM: I was a struggling assistant professor in Florida. There was no way I was going to get tenure. My dissertation did not really reflect who I was. It was too technical and it's very hard to live another academic's life including your own advisors. So what I did was to wander around the corridors of Florida and look for journal clubs and discussion meetings that were going on, work group meetings and one day I saw a paper presented by one of the graduate students which is a classic paper in case-control Studies by Prentice and Pyke. So a case-control Study is a classic design for studying rare diseases. You essentially sample the disease people who are the cases from the population and because they are rare you cannot sample the entire control population that's most of the population. So this cases are often matched but a comparable control group or a disease free group of individuals. There is a random sample taken from the disease free population as well and then you try to contrast the cases versus the controls and this is an efficient design for studying rare diseases as opposed to prospective cohort studies where this may take a long time to reach that case-control balance.
BM: That paper changed my entire academic trajectory because I just loved it. The mathematics was absolutely precise and rigorous and I felt passionate about that from my young age but also it gave credence and rigor to one way we analyzed data from case-control Studies and before that there was not a rigorous justification. So that was beautiful that everybody is going to use this tool but here is this paper, this landmark paper which is absolutely critical essential mathematics. I wanted to do something like that.
BM: I moved to University of Michigan biostatistics Department for personal reasons and this is why life is such a random walk. I moved rather begrudgingly, first of all I was moving from Florida to Michigan, the disgruntlement is obvious but then I also thought that what am I going to do? I'm going to not have any time to think about theoretical ideas. I'll always be so fragmented in terms of biomedical collaboration and I have to talk to a real investigator that really scared me and this is why I think that sometimes things which you are afraid of when you conquer your fear you can actually really make progress with a novel point of view.
BM: It turns out that University of Michigan and the School of Public Health is the best thing that could happen to me as an academic. It came to my rescue because I could see the translation of abstract mathematics into complex problems in biomedical sciences, environmental health, cancer and it has been a dream, a fairy tale, a data driven fairy tale.
Host: In 2015, Mukherjee and her colleagues in Biostatistics at Michigan Public Health started an undergraduate summer program on big data called the Big Data Summer Institute or BDSI.
BM: As an educator, I always wanted to advocate for change. Starting around 2010, we could see a lot of algorithmic modeling, machine learning AI coming into our discipline and the data sets were getting vast and enormous and heterogeneous. How do we talk about data science to undergraduates? At the time, very few universities actually had a data science undergraduate program and we started bringing in 40 undergraduates from all over the world and this program is now supported by NIH SIBS program, Summer Institute in Biostatistics by NIHLBI, has a lot of donor and philanthropic money devoted to bringing in international students have grown into a movement, a data science movement. If I ever dreamt of a data science army, this is my revolution and I train 325 soldiers who are going to change the world one algorithm or one theorem or one model at a time.
BM: More than 70% of these undergraduate students have pursued graduate school in quantitative disciplines so the program was highly influential. But most importantly, what they appreciated is the time that they had to think about the future. The impact they are going to make to humanity with their incredible talents that they have in quantitative data sciences. In our day to day undergraduate curriculum, we are so focused on our exams and homeworks and GPAs, we do not have time to reflect on the broader goal. What do I have to contribute as a minuscule existence in this vast world? And students often reflected during the program, many of them said yes, blending math and medicine into biostatistics is what I want to pursue. So I do think that this is the best thing that I have done in my academic career. I do not know whether my papers will have the longevity of time, but definitely these people who are going to train other people. This is a ripple effect and a continuum, it's definitely going to stay.
Host: In 2020, when the COVID-19 pandemic really took hold, Mukherjee knew she had to do something to help and she felt the place where she could make a difference was in India.
BM: Sometimes in life, you act out of pure madness and it may work, it may not work, but you have no other choice except doing it. And so when COVID started and I still remember it was March 16th, 2020, the first day of stay at home order in Michigan, we were all lost. Coming from India, I was very, very worried about my family, my parents and what could I do? I'm not a frontline healthcare worker. I was not working in the labs thinking about vaccines. I could help as a data scientist. At the time, my colleague Peter Song and his students were working on a new methodology, which was used to model the Wuhan outbreak. And I was getting interested in those models from a purely statistical point of view because these models are very different than the traditional linear or generalized linear models we use and teach in our courses. These are compartmental models to track the virus transmission and divide the population into different compartments and describe the flow of the epidemic. So these are very interesting non-linear models that I wanted to learn more about. But when the pandemic started to spread around the world and I saw the data and model vacuum in India, I invited my students to work on this project using that model with Dr. Song's lab's help to track and project the trajectory of the virus in India.
BM: And this was March 16th. We produced a pretty well-reasoned document and that document became viral. And this was my first foray into non-traditional medium of communication and it has tremendous traction from media. And we immediately recognized that there is a gap and a void in terms of real time forecasting of the outbreak. And we continued on that path. On March 25th, 2020, the historic national lockdown of 1.4 billion people in world's largest democracy was announced and our work was considered as a key piece of evidence that informed some of those consideration that was going on.
BM: And this journey that we started, we could have stopped it. But we realized that there are many states in India which are very small states with very limited capacity. The data and software and real-time forecasting framework was not that robust. So we created this app, COVID19.org, where we scraped data from various different sources to come up with real time projections nationally for India, but also for the states and union territories of India.
Host: So what were some of the biggest takeaways from this hectic time?
BM: First of all, to be a researcher, during the pandemic you need enormous stamina. With the time gaps in with India, I basically had two jobs. I'll finish my job at Michigan, take an evening nap and get up at midnight in order to be on meetings and seminars in India. And so the second part is really knowing the context matters. The reason that I could do a good job in modeling the pandemic and the trajectory in India as opposed to any other country in the world because I spent the first 23 years of my life and I have very strong roots to India. So I could actually understand what was going on sociopolitically in India. So the context matters.
BM: And the third thing is really stay in the game. With your objective science, if you have a presence, if you make your mistakes, they will be considered by the public as genuine mistakes, not driven by any agenda. And after doing this work for two years and 7000 media citations, I felt that we gain some respect as modelers and in contributing objectively to the pandemic. And the media work we did because of two reasons. One is that I saw very few statisticians on media. Modeling is our thing. And we saw many other people, physicists, economists, MDs, bioinformaticians talking about models that we should own in our discipline. And then the second thing was the positive of non-male voices in the media, in particularly in India. And I wanted to be a part of the process. And I also felt that we had something important to contribute to tracking the virus and also in terms of the public health prevention that were being rolled out.
Host: The success of her global work has taken her career to new heights at home as well. At the time of this recording, Mukherjee is embarking on the latest leg in her journey, as she takes on a leadership role in the office of the Vice President for Research at the University of Michigan.
BM: So que sera, sera, but I'm transitioning from the chair position in biostatistics. I was very, very honored to be the first non-male chair of biostatistics. A department which started in 1949, one of the world's finest, kindest, largest biostatistics departments. And I'm very proud to where we are at in terms of our professional reputation and how intellectually vibrant and socially progressive our community is. I have been very fortunate to have my next opportunity defined for me to really reimagine the research data services strategy across the three University of Michigan campuses. I was appointed as an assistant vice president at the office of vice president of research to think about what AI augmented brand new, massive consulting center for providing research data services may look like in the future. So I'm on a listening tour. I'm figuring out how a quantitative philosopher uses data, how a marketing researcher uses causal inference. Of course, I'm quite familiar with the data explosion of the data explosion of the biomedical and public health campus. And hopefully I'll be able to advance the university's mission of robustifying the data infrastructure and the analysis infrastructure of this magnificent research community.
Host: Mukherjee left us with a few words of encouragement for students interested in using their mathematical skills for the greater good.
BM: Advice is, I feel in some sense redundant because I received many advice and sometimes I did not follow any. So the advice really comes from your inner voice. I think that to have the faith that we are really privileged to be standing at a time where disciplinary boundaries are becoming much less rigid. You could be a professor of biostatistics or statistics and be considered a very deep expert in sociology or philosophy or cancer research or environmental health. I think it's very important to find a collaborative area where you have depth of knowledge and you are genuinely interested in answering that scientific question and the rest of the motivation and the inspiration will fall into places. It's also important to know that success takes time. And as I shared that starting as a very struggling assistant professor in Florida, I could never imagine that in the last 17 years from 2006 to 2023, I would be going through every possible rank in University of Michigan and living life, validating yourself in other people's benchmark is not worth it.
BM: At the end of the day, when you brush your teeth and I hope you do, you look at the mirror and you are only answerable to yourself and to have that confidence is hard when you are young and there is peer pressure, there is parental pressure, but please, please do honor and respect your inner voice because no one can replicate that.
Host: Thanks for listening to this episode of Population Healthy from the University of Michigan, School of Public Health, we're glad you decided to join us and hope you learned something that'll help you improve your own health or make the world a healthier place. If you enjoyed the show, please subscribe or follow this podcast on iTunes, Apple Podcasts, Google Play, Stitcher, Spotify or wherever you listen to podcasts. Be sure to follow us @umichsph on Twitter, Instagram and Facebook so you can share your perspectives on the issues we discuss, learn more from Michigan Public Health experts and share episodes of the podcast with your friends on social media. You're invited to subscribe to our weekly newsletter to get the latest research news and analysis from the University of Michigan School of Public Health. Visit publichealth.umich.edu/news/newsletter to sign up. You can also check out the show notes on our website population-healthy.com for more resources on the topics discussed in this episode. We hope you can join us for our next edition where we'll dig in further to public health topics that affect all of us at a population level.
Be sure to follow us at @umichsph on Twitter, Instagram, and Facebook, so you can share your perspectives on the issues we discussed, learn more from Michigan Public Health experts, and share episodes of the podcast with your friends on social media.
In This Episode
John D. Kalbfleisch Distinguished University Professor of Biostatistics
Assistant Vice President for Research, Office of the Vice President for Research
Bhramar Mukherjee holds prestigious positions at the University of Michigan, including John D. Kalbfleisch Distinguished University Professor and Chair of the Department of Biostatistics. She has co-authored more than 370 publications in various fields and is the founding director of the university's Big Data Summer Institute. Mukherjee and her team significantly contributed to modeling the trajectory of the SARS-CoV-2 virus in India during the COVID-19 pandemic.