Oby: From the campus of Harvard Medical School, this is Think Research, a podcast devoted to the stories behind clinical research. I'm Oby.

Brendan: And I'm Brendan, and we are your hosts. Think Research is brought to you by Harvard Catalyst, Harvard University's Clinical and Translational Science Center.

Oby: And by NCATS the National Center for Advancing Translational Sciences.

Brendan: C. difficile is the most common cause of infection in hospitals. Of people who become infected, 25% recur, which can be life threatening for those whose health is already compromised. The key to Georg Gerber's recent research is a new method of both clinically and computationally understanding prevention and recurrence.

On this episode of Think Research, Georg Gerber returns to speak with guest host Hardeep Ranu about his recent Translational Innovator Award and how his career in Hollywood shaped how he thinks about research. Dr. Georg Gerber is Associate Professor of Pathology at Brigham and Women's Hospital and Harvard Medical School.

Hardeep Ranu: Do you mind just giving a brief introduction of yourself?

Georg Gerber: Yeah, sure. So, I'm Goerg Gerber. I'm Associate Professor of Pathology at Brigham and Women's Hospital and Harvard Medical School. And my focus over the last decades has really been in computational biology. I'm a computer scientist by training, a PhD from MIT in computer science, and I also have an M.D. with training and pathology. And for about the last 10 years-- actually, I think it's been more than that now. Time goes quickly. I've been focused on the microbiome.

And it's a really fascinating area for the intersection of my interests, which are in infectious diseases and the inflammation, immunology, and the computational aspects of this extremely complicated and dynamic ecosystem that's living on and within us. It's everywhere and has these huge implications with human health.

Hardeep Ranu: OK, great. So, my first question is why did you apply for the award? What was your motivation behind it?

Georg Gerber: Yeah, so we had been doing a larger kind of clinical study on recurrence of C. difficile colitis. So, C. diff is a pathogen that causes severe diarrhea primarily in people who have taken antibiotics. And so we were trying to understand, OK, can we actually predict who's going to recur? Because there are treatment options, but they're expensive. They're not without risk, so we'd like to know. And about 25% of people who get the disease will recur with it within a span of about two months, often, actually, a lot faster.

So, we had started this study. And we were looking at the microbiome, the constituents, what the sort of composition of the bacteria. But it became evident to us we were-- the metabolic output of these bacteria is a really important factor in protection against the C. difficile recurrence, as well as potentially looking at these as biomarkers that we could use diagnostically.

So, the Catalyst opportunity was really interesting, because this was sort of a new thing we wanted to add on was looking at the metabolites as well as developing new computational methods to do that. And so it was-- in the scheme of things with research, it's a relatively small amount of money, but it was a sort of new innovative project that wasn't funded under our existing study.

The Catalyst grant was really interesting because we did attack both this clinical aspect and then a computational thing. And now those are-- nowadays, these are getting more and more tied up, practically. As you have something clinically, it's complicated, we can look at very high dimensional data and needing that computational component. So, I think the importance of the research was on the biology side. The biomedical side was really understanding how these metabolites, these small molecules that are being produced or transformed by the microbes might actually be acting to prevent the infection or recurrence of the infection.

And then on the computational side, the big question is when we have this much data and this complexity of data, how we can actually not just learn a model that predicts-- so I give you a big readout from a patient and you can tell-- have some score predicting, oh, will they recur or not, but is interpretable. And so I can actually understand how that ties into the biology. This can actually be explained clinically with what's going on.

And it turns out that is a really challenging problem. There's a lot going on in the realm of machine learning. For instance, like with image recognition, you'll see things in radiology and now in pathology where you're taking an image and it pops up with a diagnosis, but in the biomedical sphere, we want to know more. We want to know why the algorithm is making that diagnosis, what's going into it, and the so-called explainability and interpretability. And so that was the second part of the grant was developing new methods that would be interpretable.

Hardeep Ranu: Where do you think you are now, and how far away do you think you are to getting to what you were talking about, about getting to be able to print it out, having this readout and say, what are the predictability? How-- when do you see that as a possibility?

Georg Gerber: Yeah, I think it's with-- as always, when you're looking at human biology, it's very complicated. The study we had was with a relatively small number of people. It was about 75 people. But when we sort of winnowed it down, having enough samples at the right time points and whatnot, it's even smaller than that. And so we did with the study identify metabolites that are predictive of recurrence, and some of them are biologically really rational. We can make very good sense of it, and some of them are quite interesting as to what the biological implications are. So, I think we made really good progress in terms of proof of principle, showing that we can find this.

To get this to an actual clinical test, the bottom line is we'd need more people. We need to do a larger study on an independent cohort and show that this holds up, because there's a lot of variability just in metabolites in the human population. And there's also variability in the disease progression, et cetera, et cetera. So, that really would be the next step, but I think we made really exciting progress as sort of showing that, oh, these metabolites-- and the sort of punchline to it, we found, is they're a lot more predictive than the composition of the microbiome. And the way I look at that is sort of these are the end output of many different microbes that could vary a lot between people, but sort of the catalog of what they're producing is in many ways more tractable, and it's also kind of maybe the end effector in a lot of the interactions. So, that is a really promising line of research.

And on the computational side, we made really good progress with showing that we can scale this method. So, the initial method we had took about three days to run on a relatively small data set, like 100 subjects, and the current method runs in-- it's like an hour on the same scale. So, we're now able to scale practically up to studies with thousands of people. So, we're sort of at this phase where we now-- we're ready to go to the next phase, but we need additional funding to get there.

Hardeep Ranu: Presumably, that machine learning, you would-- that algorithm, you would make available to others to use it. Is that something, so it sort of benefits everybody who needs to run that kind of data and look at that kind of stuff?

Georg Gerber: Yeah, absolutely. We-- I mean, this has been the policy of my lab for day one. I release everything under open source licenses that comes out of lab. And my opinion is if it's-- I know some people differ on this, but my opinion if it's from sort of publicly funded research, I think it belongs in the public domain.

Hardeep Ranu: Yeah. So, it goes on to-- so, it's not just the sort of initial focus or the main focus is predicting C. difficile reoccurrence, but it's also here's this whole algorithm that could be used in something completely different. So, I think that's really a bonus, if you like, to this kind of a project, that it can also be used elsewhere, and not just within this sort of narrow field, if you want to call it that.

Georg Gerber: Yeah. I mean, it's-- I absolutely think so. I mean, there's-- one of the problems is it's difficult under the NIH model to get funding to develop tools and technology, because the instances are so disease-focused. The NIGMS is the one exception, and that's actually where I have an NIH grant where you can do more tool building, but it's a huge thing. I mean, look at the extent to which computational tools, or in some cases experimental tools have just enabled so many fields of research, things that weren't even imagined when they were initially developed.

So, yeah, we have a paper that I keep hoping is going to be finished, but will be soon, that presents the method. We have-- the source code is going to be on GitHub. And we've gone further than that. We actually made an application that's easy to use. And in the paper, we test it on a whole bunch of different data sets. You know, it's not just our data.

Hardeep Ranu: What do you use the money for? $50,000, like you said. And we all know that it's not really that much money. Especially, it seems now in some ways, it almost feels trivial versus 10 years ago, when it would have gone a lot further but how did you use it? How did you maximize those $50,000 for the most gain you could get?

Georg Gerber: Yeah, so you know, this supported primarily a new student in my lab, honestly. There was some support also to other staff in my lab, but it's interesting on the computational side, but you know even on the experimental side in a lot of ways, the biggest expense, of course, is people. And so it seems like a trivial amount of money in some ways, but this partially supported her. She got going in the lab. She's now on her way to doing-- this will be part of her thesis. So, I mean, in my opinion, that, to me, if I have money, that's sort of where I feel I can spend it best is investing in a young investigator's career.

A lot of these big-- I'm going to say this, and it may be an unpopular thing to say-- but a lot of these big experimental data collection expeditions don't necessarily result in that much. They're very, very expensive. Whereas training someone and actually producing it, that almost always is a worthwhile investment.

Hardeep Ranu: Right. Yeah. Yeah, it's a really good point. And I think that it's not something that people think about in terms of research.

Georg Gerber: Yeah, it's really critical. And in the machine learning realm, it's increasingly challenging, because of course our competitor now is not just other academic labs, but in many ways, it's primarily industry. And you know, what they can offer in terms of salary is so much greater than what we can ever do academically. So, I think why people stay or will put energy into it is a combination of the projects are really interesting, they feel it's something that will make an impact on society, and they feel like they're really being mentored and going somewhere with their career intellectually. So, that's kind of what we can offer in the academic setting, but it's challenging, right now, with the Googles and Amazons of the world.

Hardeep Ranu: Yeah, yeah, yeah, undoubtedly. But like you said, it's sort of one of those things where if that person is going for the money, that's one thing. But also, what do they value in their work? At the end of the day, what is it? Is it sort of-- is it making money and just working for Google and getting all the perks of there versus let's say the intellectual freedom when working in an academic lab, or that there's an opportunity to pursue slightly different things, if you wanted to, in academia. I mean, I know nothing about what it's like to work as a programmer at Google or Amazon, but it would seem that would be the case.

Georg Gerber: Yeah, I mean, it's-- I think there's always been this tension in engineering that's different than, say, the biological sciences, because there are sort of straight out of-- even out of just a bachelor's degree, there are good career opportunities in engineering, and some of them are quite interesting. I mean, many of these companies have very robust research labs where they really are innovating, so it's not like just an engineering programming kind of job, although honestly, the vast majority of the jobs are that.

And so it is challenging. And I think the one thing I would say that we have that they don't is the biomedical piece. I mean, you're not going to be at Google, you're not going to be seeing patients and interacting on that level. So, it's-- I think for people where that's where their passion is, this is where they're going to end up.

I think the other thing we need to be cognizant of as a field, I run into this a lot every day, is yes, I think people are willing to work for less to have some of what we talked about, freedom, feeling like they're making important contributions. But when that gap reaches a certain level where you can't support your family or really giving up just too much, you start to see a lot of leakage. And that's happened a lot now, because I think in part just the tech area, particularly in deep learning, has just accelerated so fast and what's being offered in the industry is orders of magnitude more.

Hardeep Ranu: Yeah. So, yeah, at some point, if you're being paid-- and I'm making it up-- $50,000, let's say, a year versus going to Google and being paid, I don't know, 3 times, 4 times that, at some point, you kind of think, what are the benefits of going to somewhere where you want to be able to afford to buy a house, especially in Boston, I think?

Georg Gerber: Right, right. Yeah, yeah. And you know, it's all the-- I mean, Boston, San Francisco, New York, I mean, a lot of the metro areas, just the cost of living is so high. And it's a challenge. This is-- people ask me sort of what the biggest challenge is, and this is the biggest challenge in the field right now is getting good people and competing with the options that are out there right now.

Hardeep Ranu: In terms of the model that we have within Translational Innovator is to meet with the investigators every four to six weeks, and then we get that particular cohort of awardees together every four months, what kind of value in that did you find?

Georg Gerber: I think it's a great idea. I mean, it fell apart a tiny bit because of COVID right near the end. But when it was happening, I think it was great. I mean, I think it serves a few functions. Some funding institutions I've dealt with, you kind of get the feeling that they really don't care about you and they're dumping the money in your lap or sometimes maybe not, you get the feeling they're going to pull the money away. I feel like it's very anonymous feeling.

So, I think that's helpful. You really feel that Catalyst wants the projects to succeed. And when we'd meet, you were always asking what can I do here if you have obstacles. So, I think that's really-- it feels really good, is really helpful. But then I think the other aspect was definitely meeting with the other investigators was really nice because you see-- the awards are thematic, so everything was on microbiome.

And I was seeing these other projects and what people's ideas are. And one of the worries worked a lot on viruses in the gut, and that was really interesting to sort of have a back and forth with her because that's not my area, but it's one I'm very interested in. We've actually done a little bit of work in that area. And others, I know, were picking my brain about computational things. So, I think it's a great forum to exchange ideas and help advance the field overall.

Hardeep Ranu: Yeah, that's what I've heard, as well, from other awardees, that it's sort of the meetings with the other awardees where it sort of helped with maybe expanding what they were doing or maybe thinking about other ways which their research could go or adding on to what they were doing already. It certainly seems that that is something that people have valued. And from a Catalyst point of view, that's exactly what we want is we want people from different areas, different backgrounds to be able to connect with others in order to push their research forward or into a different direction.

Georg Gerber: Yeah, and I think particularly at a place like Harvard, it's really valuable, because we're very siloed. And I primarily interact with people, honestly, in my own department or on the quad. But if they're at, say, MGH or other places, I don't interact with them that much. So, it's really good to start breaking down some of the silos.

Hardeep Ranu: So, I have just been-- this is sort of separate, but it got me thinking. I had just finished reading this book called Range by David Epstein, and he makes the analogy between the Tiger Woods of the world and the Roger Federers, so Tiger Woods playing golf when he was 18 months old versus Roger Federer, who did all these other sports and then came to tennis relatively late, but that both people have ended up top of their sports. In their careers, they've accomplished a lot.

So, one of the things he talks about is the fact that people who don't necessarily start off in the profession that they land in, where they feel the most comfortable, is because they've tried all these other things, and that also the innovation can come from-- because they've had this other experience in some other field, that they can apply that to what they're currently working in. And it actually made me think of you and your background and graphics and working in Hollywood, which is completely different to what you're doing now, I would imagine.

But how much do you think that has helped you-- I know sort of on the computational side of things, obviously. But aside from that, what else do you think has helped to you? Because you made a complete career switch, correct?

Georg Gerber: Yeah, it was pretty complete.

Hardeep Ranu: Right, yeah.

Georg Gerber: I mean, I was previously making movies and had moved into even sort of management roles with that. Yeah, I mean, I think it's a really interesting point. I mean, I'll just make a brief segue because I-- in the same way of feeling a bit tangential, then I'll come back to your direct question. But one of the interesting things during COVID that I got involved with in part because there were circumstances going on in the world that happened at the same time, but also in part because it was sort of possible for me to do that was the diversity, equity, and inclusion efforts in our department. So, I chaired a task force looking at those issues. And we developed a report. And now actually following up, I think we're finally going to have someone to take some of that on and start implementing it.

But I think the issue you raise is really interesting is sort of people coming in with ultimately a different perspective. And some of that can come from they were in different careers. Some of it can come from they grew up in a different country, a different culture, all these aspects. And I think it makes a huge difference in a field when you're not only sort of tolerating that but actually encouraging it.

Because I can't tell you, you look at a lot of the staff in any given field, and it can be very homogeneous and everyone is kind of working on the same problems and trying to do things the same way. And I suppose progress will be made that way, but it's a certain type of progress, and I think it's missing a lot of directions things could go in. And frankly, at Harvard, it is something we continue to lack. People will look at a candidate, and I think some of the criteria people are trying to use are pretty superficial as to what is going to be diverse. But I think things, for instance, like someone's economic background and where they came from is really, really important. We need to turn to these things and look at the bigger picture.

As far as my own path and what has been valuable, yeah, certainly, there is a piece, as far as the computer science piece, as you mentioned. But another thing that's been really interesting and it comes up a lot is one of the things you do in Hollywood is essentially telling stories and presenting things in a way, be it information or entertainment, or whatever, in a way that the public-- it hooks the public in. And science is not entirely different. In fact, I feel like increasingly this is becoming important, recognized as important, as your clients, so to speak, aren't only other scientists, but it's the public.

So, this is a skill I did learn my other job is sort of how to make these things compelling and how to communicate them. It's marketing on a certain level. It's storytelling on another level. I mean, I can't tell you the number of papers I see where I think there's a good idea there, but they can't or haven't communicated in a way where I understand what the innovation, what's interesting, what's compelling. And this-- you see this increasingly with funding agencies, because they have to go out there. And often, like NIH, they have to convince Congress that what they're doing is important and relevant.

And it can't just be highly technical stuff, but also, I think they have this tendency everything has to translate into oh, we're curing a disease. And I think they actually need to spend more time and energy communicating why some of these technology developments are exciting, like mRNA vaccines. Before COVID, hardly anyone had heard of them. I mean, it was and is a very exciting technology and really potentially groundbreaking, but these things to the average person are not well communicated.

Hardeep Ranu: Right, right. I mean now nowadays people are like, oh, you need to get the PCR test. And it's like-- it's like, never would I have thought people would be saying the letters PCR together, never mind saying, oh well, that's the gold standard for testing, or you know.

Georg Gerber: Yeah, yeah, yeah. No, and I mean, it also plays into the broader issues we were talking about before with education in the population. I mean, people are understanding these things have-- it has an impact. It's important, right? I think when there's misunderstanding and mistrust of what's going on, it leads to some really irrational and weird behavior and decisions of the public. I've heard the things that-- you know, people are convinced the PCR tests are doing all sorts of strange things. They think the vaccine is magnetized and tracking you.

These are things that on the one hand, it sounds very silly to a scientist, but on the other hand, I appreciate people-- they really haven't been informed about these things, and so they're just sort of left on their own to speculate. Or I mean, I've read some of the articles that are out there for the lay public, and they're not very well written. I can't-- even knowing what the technology is, I can't actually piece together how it maps onto what's actually going on

Hardeep Ranu: Yeah, yeah, yeah. I think one of the things is this sort of use of jargon within the science, that it's actually not necessary at all, that in fact, actually, a clear, concise explanation of what it is that someone is researching goes so much further than just having a whole load of jargon in there.

Georg Gerber: Yeah, no. I actually-- and I mean, I really emphasize this to students. I actually think usually when I read something loaded with jargon, the person doesn't know what they're talking about. It's resorting to a lot of sort of shorthand of things I don't think they understand.

Some of the funding agencies, I think NSF has sort of a requirement like this where you have to write these abstracts that are supposed to be jargon free. I mean, I don't know-- the problem is I don't know who's reviewing them, because I've read some that are certainly not jargon free. But, I tried to do it when I wrote mine. I mean, it's sort of the idea of can you write this on a fifth grade reading level. And you can write a lot of things on a fifth grade reading level that are-- convey very complicated information. It's just a matter you sort of have to go through it step by step and as you said not using jargon and all that.

Hardeep Ranu: And especially, I think, for the Catalyst awards, none of us looking at them are experts in those fields, so it's difficult to then assess that it might be something that is-- like you said, that it is really, really interesting, really novel, but because there's all this jargon there, you can't really quite grasp the concept behind it.

Georg Gerber: This is the other thing I counsel students on is the person reviewing a paper or a grant is supposedly an expert in the field, but it comes down to what I'd call a theory of mind issue, to use some jargon there. But the idea is, can you get inside someone else's head? And they're assuming that the reviewer or reader has sort of omniscient capabilities, right, that they know everything about the project and every paper they've read and everything, which is ridiculous. I mean, even someone who's an expert in the field is not going to be able to follow something that is just loaded with jargon and lots of unnecessary technical details.

Hardeep Ranu: That was great. Thanks for taking the time.

[music playing]

Brendan: Thank you for listening. If you've enjoyed this podcast, please rate us on iTunes and help us spread the word about the amazing research taking place across the Harvard community.

Oby: To learn more about the guests on this episode, visit our website, catalyst.harvard.edu/thinkresearch.