Race & Genetics in America

Transcript for Student Voices

Lorin Crawford/Chib Nwizu

[music]

You’re listening to Student Voices, a podcast featuring student-led interviews of Brown University faculty based on the Race & in America panel discussion series, curated by the Center for the Study of Race and Ethnicity in America in partnership with the Office of the Provost.

Chib Nwizu: I’m Chibuikem Nwizu, I am an MD-PhD candidate in the Department of Computational Molecular Biology and I'm speaking with Dr. Lorin Crawford about the recent panel discussion on Race & Genetics in America.

Professor Lorin Crawford: Cool. Thanks for having me.

Chib Nwizu: Yeah. So in the panel discussion of Race & Genetics in America one thing that really stood out to me was the power and the potential of data science and statistics to really help us start to answer some of these really hard questions. And one thing that you mentioned is the ability of statistics and data science and development of methods in this space to really make some of these concepts more concrete. One thing, though, that I had a question about was these definitions of who belongs to what population, who belongs to what group. They seem to be defined either by a governing body that may have their own special interest or by the individual who, rightfully so, has the right to say, “I belong to this population, I belong to this other population.”

How do you as a statistician [and] a data scientist think that that affects methods development, which if the definition of who belongs to what group is fluid, how do you develop methods to kind of deal with and handle that problem?

Professor Lorin Crawford: Yeah, I think it's important for us to separate the social aspects of some of these binnings and the true biological definitions that actually contribute to health and genetics and actually have something to do with again, biology. So for instance there's a lot of conversations right now which are, you know, thinking about [how] self reported, self identified ethnicity is a box that people check, right? And I think Sohini [Ramachandran] mentioned this during the panel discussion, which I think is very on point, which is we like to disparatize people and that's a whole ‘nother issue, right? We say you're African and you're like, well, Africa's a large place. And you’re in one bin, but there are places where some of those self reportings can be helpful and one of those areas is if we think about add mixtures. So let's say you have a self-reporting file and you're saying “I'm African American and I also identify as being Latin American” or something like that. That gives some type of evidence that maybe when we're analyzing your particular genome we should be looking for things like add mixture or mixed ancestry. And so I think the key of what you're saying and what you kind of mentioned in your question is there's a difference between using some of these social indicators that people kind of identify with and then what we actually say we’re reporting on in terms of after we run models. There's a difference, there’s a responsibility of us as data scientists to make sure that we are truly being pure in our intentions about this is what we mean by ancestry and this is not what we're saying about racial and ethnic groups. We want to be very careful about that. So even if we're using that as evidence of mixed ancestry, if I then find enrichment of snips or genes or mechanisms downstream, we want to be clear that that was based on this idea of mixed ancestry and not this person being identified as these racial groups.

And so I think what our field needs to do, and you kind of mentioned this as well, is there's no real standardization of language. I use this language here maybe because it's easy for me, I don't have to really think about the difference between ancestral lineage and other social constructs. Maybe these groups are the torchbearers in our space and this is what they use and so I also need to use what they use. I think what we need to do -- and I mentioned this in the panel as data scientists, statisticians, computer scientists, whatever -- is we really need to think about working cross cuttingly with people in the social sciences, people who really think about language. And as method developers really think about 1) how to incorporate more concretely standardized language into our models, but also into our analyses, also into our reporting of those results and everything downstream after we build methods. And so I think [as] I was saying to you a little bit earlier, the language plays a huge role in other areas of sciences and I think here we need to really take a secondary look at its effects on what we are thinking about in terms of this idea and these fallacies that are being bred from race in genetics.

Chib Nwizu: Yeah. I really like that and kind of going off of that, working with social scientists seems like the most obvious place to begin, but I guess the beautiful thing about languages is it is fluid. But if we're thinking concretely in terms that do not switch if we change from person to person, how do you as a method developer, can you maybe talk about how you would envision hearing a definition from a group of social scientists and then making that both concise enough that it means the same thing but also incorporates the nuances of the human experience?

Professor Lorin Crawford: Yeah. That's definitely a tough question. The easiest part to answer that question is when you're reporting something, “say what you mean, mean what you say” kind of thing, right? Like if you mean ancestry, say that. Don't say anything about ethnicity if that's something you do not mean. Don't say these are polygenic risk scores, these polygenic risk scores don't transfer over to Blacks. Say exactly what you actually mean. If you're reporting on ancestry, let's stick to ancestry, let's not move outside of that. Places in the social sciences do have more concrete things and I think the way that you move them over to our space as modelers is being very careful and making sure that whatever analyses you were doing in whatever context you were doing it, let's stick to our inferences about that context and have it be localized only in that space. But you're absolutely right.

Look, these definitions of the analyses are not, and this is no pun intended, these are not black and white types of things, right? This is something where there's some spatial temporal gene by inner environmental interactions at play, and these are much more convoluted types of analyses that we're doing. And it's not, you know, especially when you're thinking about genetic architecture and things like that, there are many, many influencers. And one thing that I've been thinking about a lot with people, both at Brown and in Microsoft Research [New England] in different contexts, is thinking about for what phenotypes are certain binnings more appropriate than others? So you can think about something like a sickle cell disease, or maybe you want to think about different ancestries, but there also might be things like educational attainment or other things where social aspects might be much more important [to] binning people in certain ways than thinking about their ancestral background. Right? So maybe I want to think about things like zip codes or something like that as a way to kind of understand what's happening in terms of variation of that trait. I think we just need to be more thoughtful and maybe put more effort towards research in that direction of experimental design before we even start to think about the modeling aspect of stuff.

Chib Nwizu: That is really cool and it kind of leads to another question that I've had. You know, you as an expert, you get to see kind of what the data and what the models can say, and what the data and the models can't say. And knowing the limit of some of the claims that can be made I guess the question is twofold. Do you see certain claims that may be impossible to make about race from genetic information that can be made, that have been made or may be made? And I guess the second part to that and maybe the easier part is how do you think as an expert you can help port some of these really complicated ideas to someone who isn't an expert?

Professor Lorin Crawford: Right. So I think within the genetics community we do a really good job of making sure we each don't overstate things. What I think we continually need to do is have more outward facing conversations about this with other people. This idea of in layman's terms this is what we are studying and this is what we actually mean in this particular study. So this is what we mean when we're trying to predict height or whatever in these individuals and this is what the common, you know, find some kind of common language. I think that idea of not working so much in silos about what it is that people who do statistical genetics, population genetics, are actually doing. I think that conversations like this, like the one that we just had, probably need to be had more because there are all these misconceptions and you can easily see where they come from because we ourselves sometimes use these things interchangeably in papers, and headlines are made from some of these papers and that's not exactly what certain studies are saying. So I think this idea of having more conversations like the one that Tricia Rose just set up, having that be more frequent with the general public. It's tough. This is a tough conversation to have 1) because there's not really one amazing solution. What I find happens a lot when I talk to people like that, you suggest one thing and then that opens a can of worms for a ton of other problems. This is a tough, this is a very, very tough topic, but I think if we continue to have that conversation and continually try to chip away at this and also start having this conversation with the general public so that we normalize what we see as being massive differences, I think you could start chipping away at some of these fallacies. It's just like what Brandon said, it's our job as scientists to make a racist person’s job a lot harder, right? Don't use what we're doing as tools for your particular agenda. And I think these are things that we can continue to do as a community in order to continue to get rid of some of these common misconceptions.

Chib Nwizu: That's really good. I guess another question that I have and something that I've been thinking about a lot is, you know, we like to tout science, technology, math, and statistics as infallible areas of study and there isn't a gray area, but I think that the reality is we have to recognize that as a scientist, as a modeler, we have our own set of biases that we kind of bring into the modeling process. And in a space that for the most part has been dominated by individuals who, you know, come from a very particular background, straight white males, how do you see the inclusion of more people? Like you yourself are Black and so you kind of have a unique experience that you bring into the modeling experience. Do you think that there also needs to be an increase in diversity amongst statistician modelers?

Professor Lorin Crawford: That's my own personal community. Absolutely, yes. Computational biology, absolutely. You see in our lab, our lab group I think is what science should look like, right? Diversity of people, it brings this beautiful result to the work that we do. And yeah, absolutely, I think particularly for this particular topic, yes. It wasn't just cool that we got to have that conversation with Tricia Rose, it was also the people who got to have the conversation with Tricia Rose, right? I think that was the coolest part about it and I hope that we continue to, I mean, there are people across campus like Emilia Huerta-Sanchez and other people who would make amazing contributions, to hearing even what she has to say about her work. In short we absolutely need to do a better job of not only being inclusive like I was saying before, like who we work with, but also who we're training, who we're giving opportunities to and who we're hearing from on this particular topic because I think this is, like I said before, this isn't going to be an easy fix. This isn't something that us moving forward we're going to have a direct path to. I think it's going to take a lot of people coming at this from many different angles moving forward.

Chib Nwizu: Yeah. And you said something earlier that I found to be really powerful and I wanted to hear you say it again, that we as a community also have to be more open to being wrong and being able to correct others.

Professor Lorin Crawford: The only reason why I really started digging even deeper into this is [because of] Mary Gray, an anthropologist by training at Microsoft Research. I'll never forget this, I gave a talk on some stuff Sohini and I were working on and she was like, “I think that's wrong. I think what you're saying is not right. You can't mean this as a way of studying biology.” She wass like, “You can't mean that.” And just by that, just that little spark of, “Oh, okay. Send me books. Let me learn.” You know, because yes, I identify as being African American, but I don't know half of what I should probably or need to know on this topic.

And so I think, yes, including people who have different expertise and who are going to be able to enrich your thought process on these things and allow you to do your own work in your own discipline better and more thoughtfully, that's what this is all about. So, yeah, that's where I come from. You have to be okay with being wrong and this is a tough subject to put yourself out there and being wrong on or maybe saying the wrong thing and that kind of thing. But I think it's important. This is something that we have to continue to do.

Chib Nwizu: Yeah. Thank you so much for having this conversation, really being a participant in this conversation about race and genetics and thank you so much, Dr. Crawford. Can’t wait to see what work this conversation leads to.

Professor Lorin Crawford: Yes, absolutely. Thanks for having me.

[music]

Student Voices is a feature of the Race & in America digital publication series developed by the Brown University Library. Our theme music is “see the unseen” by Butter. Explore the series at DigitalPublications.Brown.edu