Recreating Faces of Egyptian Mummies Using Ancient DNA

In this interview, Ellen Greytak (Director of Bioinformatics at Parabon Nanolabs) discusses how DNA advances have changed the field of anthropology and what we know about ancient humans.


She also describes her poster presented at the 32nd International Symposium on Human Identification in which she shared the predicted faces of three ancient mummies from an ancient Nile community in Egypt known as the Abusir el-Meleq.


DNA samples samples from the ancient mummies, estimated to be between 2,023 and 2,797 years old, were processed by researchers from the Max Planck Institute for the Science of Human History and the University of Tubingen in Germany (Schuenemann et al. 2017).


Researchers at Parabon NanoLabs, used that genetic data to create 3D models of the mummies’ faces through a process called forensic DNA phenotyping, which uses genetic analysis to predict the shape of facial features and other aspects of a person’s physical appearance. This is the first time comprehensive DNA phenotyping has been performed on human DNA of this age.


To learn more about forensic DNA phenotyping, watch our Phenotyping playlist on our YouTube channel.





Laura: Thank you everyone for joining us for our annual video series for the International Symposium on Human Identification. We have a special guest with us today. Ellen, why don’t you tell us a little bit about yourself and your work?


Ellen: Hi, I’m Ellen Greytak. I’m the Director of Bioinformatics at Parabon Nanonlabs. These days, I think we’re mostly known for doing genetic genealogy, which is awesome, but we do much more than that. We started with DNA phenotyping, and we also work in medical areas in non-forensics, so we do a lot of different projects, but the general theme of our work is: given a DNA sample, what can we say about that organism?


Laura: Wonderful, thank you. Well, why don’t we talk a little bit about the poster you submitted for this year’s symposium. What was that?

Ellen: The poster I did this year is called DNA Phenotyping Using Ancient DNA from Ancient Egyptian Mummies. So, this idea came about, because we found that there’s a lot of data freely available online from ancient sequencing projects that have been going on, and it was sort of a question of, could we get the data to an appropriate level where we could do DNA phenotyping on it? On someone from thousands of years ago? So, it was sort of a proof of concept that these same tools that are being used for ancient DNA can be used in forensics where we have very difficult samples as well.


Laura: That’s fascinating! How do you use the techniques that we have for DNA to gather the genomes of people who lived thousands of years ago?


Ellen: Yeah, the whole field of archeo-genomics is really fascinating and has come a long way since the neanderthal genome was first published. Now there are thousands and thousands of ancient sequences out there that have been published, so there have been innovations both on the laboratory side and on the bioinformatics side.

On the laboratory side, there have been developments in capture assays, where a big problem with ancient DNA is that when you extract that sample, most of the DNA you find is actually modern bacteria, and that DNA is in much better shape than the ancient human, whose DNA you’re trying to analyze. So, these capture assays have baits that actually pull out the sequences that belong to the humans, so that way, you can enrich for the sequences that you’re interested in. You’re still going to be sequencing some bacteria, but you can pull that human proportion up from below 1% to several percent or 10’s of percents.

In addition, on the bioinformatics side, being able to go from these low coverage genomes to really high quality data, that’s something that’s just come about in the last year or two. I’m sure you’ve heard Ed Green talk about it. This low coverage amputation. The idea that we can go from really low coverage genomes. The DNA that we had in this case, they were all far less than 1x coverage, and yet we were able to apply low coverage amputation to recover most of the SNPs that we needed for DNA phenotyping. So, we basically take something that’s covered at less than 1x and get hundreds of thousands of pretty accurate SNPs from that. So, like I said, this was sort of a proof of concept in this cool idea with the ancient Egyptian mummies, but those same techniques are also being used in our forensics cases.


Laura: That’s fascinating. So, now that you’re able to gather this high-quality data, what does that mean for people who are studying ancient cultures or maybe also what that means for those using forensic DNA for other purposes?


Ellen: Well, I think ancient DNA has really changed the face of anthropology, because, for a long time, the only way that we could learn about how cultures interacted was trying to infer from what we found. Now, we can directly see how the DNA flowed from one place to another in different time periods. It’s been really fascinating to see. I feel like every day there’s been a new paper coming out with new ancient genomes comparing them to the modern populations that are there and being able to tell the demographic story of the human population.


Laura: So what are some of the challenges that you’ve faced while doing this. Every year, we take a leap forward, but not without a lot of hard work. What have you learned along the way?


Ellen: Well, these data sets are really, really tough, and definitely a few years ago, we were of the same opinion of everyone else, where if you have less than 1x coverage, there really wasn’t a lot that you could do. You could call a few SNPs here and there where you got lucky, but you really couldn’t put together a high-quality genome-wide covered set of SNPs, but now we know that it is possible thanks to research that’s been done and is out there, so it’s just a matter of implementing that pipeline. But, there’s only so much that you can do. With the poster that I’m showing, there are 3 samples. One of them had close to 1x coverage, and that we were able to get beautiful data out of. The others were much lower, so we were not able to impute nearly as many SNPs, so we had to do a little more guesswork on the phenotypes, which means we have much bigger confidence intervals, so it’s a good use case on what we can and can’t do with these types of data.


Laura: Talking about the phenotypes, what predictions were you able to make in this particular case? What did you find?


Ellen: Pretty neat. In the original paper, where they had published these original sequences, they had looked for a known SNP for fair skin color in the higher coverage sample, and they found that it was there, so they thought that he probably had lighter skin, but that was the extent of that. We did find that as well, but unfortunately, like I mentioned, we were not able to get all of the SNPs that we were interested in. There is one particular SNP that is very, very important for eye color and we were not able get a genotype for it in the two lower coverage samples. So, basically, I made predictions for them assuming multiple different genotypes. So, they did all come out as fair skinned, brown hair, and dark eyes. I think the ancestry was the most interesting part. They really did not cluster with modern Egyptians, but instead with people more from the Arabian Peninsula, more up into the Middle East rather than Africa. So, that’s also what’s been found in the paper. There really was no sign of sub-Saharan African ancestry in these individuals. So, that must have come in during a later phase of the Egyptian population.


Laura: It’s fascinating that you can confirm some of these theories or add new evidence to the case. That’s really amazing.


Ellen: Yeah, it’s very cool. We’re now looking for what other genomes are out there. Genetic genealogy for ancient samples, we really wouldn’t expect to be able to find anything, but these same techniques, we’ve been using them in our casework, and we are able to do genetic genealogy on really difficult bone samples where there’s a lot of microbial contamination, not a lot of coverage on the genome, and we’re getting beautiful, nicely imputed data sets. So, this poster is also showing a validation on the amputation process and showing how we can go from really low coverage to really nice data.


Laura: That’s really remarkable. How did you get involved in this work in the first place? I always like to ask that.


Ellen: Well, I did my PhD in evolutionary biology, and after my post-doc, I was just looking for a job and happened to see this posting where they had a grant to predict what people look like from DNA, and they were looking for a biologist to lead it. I thought that sounded like just about the coolest thing that I could imagine, so I was really very, very fortunate that Parabon was looking for someone when I was looking for a job, because it’s been so much fun. Every day I get to learn something new, so I love my work.


Laura: Any recommendations for people who might be thinking about entering the field?


Ellen: For me, the biggest change… So obviously things have changed a bit in the past 10 years since I’ve graduated but learning how to program changed everything. When I got out of grad school, I could not find a job. No one really wanted an evolutionary biologist, but once I learned how to program and it became not just “I can answer some questions”, but “I can write something. I can create something new to ask new questions”, all of a sudden everything opened up. There were so many options out there, so I always tell anybody if you’re looking for one skill to work on, learn how to program.


Laura: That’s a good tip. I don’t know that we’ve had that tip before. It could help a lot of people. Can you talk about what you’re working on now?


Ellen: Well, we’re working on lots of different things. Of course, we always have the casework going on. Right now, the big project that I’m working on is studying antibiotic resistance in bacteria. So, analyzing the genome of a bacterial strain and predicting what antibiotics it would be resistant to, so that’s what I’ve been up to these days.


Laura: Wow, that’s fascinating, especially in these days in light of everything we’ve just been going through. Absolutely, and the reason that we’re digital as people can see for a little bit longer. Anything else you want to talk about, like maybe what the future holds for phenotyping given what you’ve seen and the progress over the last couple of years?


Ellen: The literature has been full of really cool papers, both expanding the populations for phenotyping and expanding the types of phenotypes that we can do. I’m still really excited about the possibility of predicting age from methylation. I feel like I’ve been saying that for 5 years now, but I’m really hoping that one day it will start actually being able to be used in forensic cases. That would really make a huge difference not only for phenotyping, but also for genetic genealogy. Knowing the generation of the person that you’re looking for instead of it’s got to be a person between 14 and 54 or something like that could make a huge difference in some of these cases.


Laura: Absolutely. I’ll look forward to talking about that in the future. Ellen, thank you so much. We really appreciate you taking the time and I hope to see you at ISHI in person in the not too distant future.


Ellen: Me too, thank you!