It has been nearly two decades since scientists accomplished the first complete sequencing of the human genome. This historic moment gave us an unprecedented view of human DNA, the genetic code that determines everything from our eye color to our chance of disease, unlocking some of the biggest mysteries of human life.
Twenty years later, despite the prevalence of genetic sequencing, considerable work remains to fulfill the promise of these advances to alleviate and cure human illness and disease.
Scientists and researchers are “actually extremely good at reading genomes, but we’re very, very bad at understanding what we’re reading,” said Lea Starita, co-director of Brotman Baty Institute for Precision Medicine’s Advanced Technology Lab, and research assistant professor in the Department of Genome Sciences at the University of Washington.
But that is changing thanks to new tools and approaches, including one called Deep Mutational Scanning. This powerful technique for determining genetic variants is generating widespread interest in the field of genetics and personalized medicine, and it’s the subject of a symposium and workshop on Jan. 13 and 14 at the University of Washington.
“I think approaches like Deep Mutational Scanning will eventually allow us to make better countermeasures, both vaccines and drugs that will help us combat even these viruses that are changing very rapidly” said Jesse Bloom, an evolutionary and computational biologist at the Fred Hutchinson Cancer Research Center, the Howard Hughes Medical Institute and the University of Washington Department of Genome Sciences.
Bloom, who researches the evolution of viruses, will deliver the keynote at the symposium, held by the Brotman Baty Institute and the Center for the Multiplex Assessment of Phenotype.
On this episode of the GeekWire Health Tech Podcast, we get a preview and a deeper understanding of Deep Mutational Scanning from Bloom and Starita.
Listen to the episode above, or subscribe in your favorite podcast app, and continue reading for an edited transcript.
Todd Bishop: Let’s start with the landscape for precision medicine and personalized medicine. Can you give us a layperson’s understanding of how personalized medicine differs from the medicine that most of us have encountered in our lives?
Lea Starita: One of the goals of precision medicine is to use the genomic sequence, the DNA sequence of the human in front of the doctor, to inform the best course of action that would be tailored to that person given their set of genes and the mutations within them.
TB: Some people in general might respond to certain treatments in certain ways and others might not. Today we don’t know necessarily why that’s the case, but personalized medicine is a quest to tailor the treatment or …
Starita: To the individual. Exactly. That’s kind of personalized medicine, but you could also extend that to infectious disease to make sure that you’re actually treating the pathogen that the person has, not the general pathogen, if you would. How would you say that, Jesse?
Jesse Bloom: I would elaborate on what Lea said when it comes to infectious diseases and other diseases. Not everybody gets equally sick when they are afflicted with the same underlying thing, and people tend to respond very differently to treatments. That obviously goes for genetic diseases caused by changes in our own genes like cancer, and it also happens with infectious diseases. For instance, the flu virus. Different people will get flu in the same year and some of them will get sicker than others, and that’s personalized variation. Obviously we’d like to be able to understand what the basis of that variation is and why some people get more sick in some years than others.
TB: Where are we today as a society, as a world, in the evolution of personalized medicine?
Starita: Pretty close to the starting line still. There’s been revolutions in DNA sequencing, for example. We’ve got a thousand dollar genome, right? So we’re actually extremely good at reading genomes, but we’re very, very bad at understanding what we’re reading. So you could imagine you’ve got a human genome, it’s three billion base pairs times two, because you’ve got two copies of your genome, one from your mother, one from your father, and within that there’s going to be millions of changes, little spelling mistakes all over the genome. We are right now very, very, very — I can’t even use enough “very”s — bad at predicting which ones of those spelling mistakes are going to either be associated with disease or predictive of disease, even for genes where we know a lot about it. Even if that spelling mistake is in a spot in the genome we know a lot about, say breast cancer genes or something like that, we are still extraordinarily bad at understanding or predicting what effects those changes might have on health.
Bloom: In our research, we’re obviously also interested in how the genetics of a person influences how sick they get with an infectious disease, but we especially focus on the fact that the viruses themselves are changing a lot, as well. So there’s changes in the virus as well as the fact that we’re all genetically different and those will interact with each other. In both cases, it really comes back to what Lea is saying is that I think we’ve reached the point in a lot of these fields where we can now determine the sequences of a human’s genome or we can determine the sequence of a virus’ genome relatively easily. But it’s still very hard to understand what those changes mean. And so, that’s really the goal of what we’re trying to do.
TB: What is deep mutational scanning in this context?
Lea Starita: A mutation is a change in the DNA sequence. DNA is just As, Cs, Ts and Gs. Some mutations which are called variants are harmless. You can think of a spelling mistake or a difference in spelling that wouldn’t change the word, right? So the American gray, which is G-R-A-Y versus the British grey, G-R-E-Y. If you saw that in a sentence, it’s gray. It’s the color.
But then it could be a spelling mistake that completely blows up the function of a protein, and then in that case, somebody could have a terrible genetic disease or could have an extremely high risk of cancer, or a flu virus could now be resistant to a drug or something like that, or resistant to your immune response. Or, mutations could also be beneficial, right? This is what allows evolution. This is how flu viruses of all the bacteria evolve to become drug resistant or gain some new enzymatic function that it needs to survive.
Bloom: For instance, in the case of mutations in the human genome, we know that everybody has mutations relative to the average human. Some of those mutations will have really major effects, some of them won’t. The very traditional way — or the way that people have first tried to understand what those mutations do — is to sequence the genomes of a group of people and then compare them. Maybe here are people who got cancer and here are people who didn’t get cancer and now you look to see which mutations are in the group that got cancer versus the group that didn’t, and you’ll try to hypothesize that the mutations that are enriched in the group that did get cancer are associated with causing cancer.
This is a really powerful approach, but it comes with a shortcoming which is that there’s a lot of mutations, and it gets very expensive to look across very, very large groups of people. And so the idea of a technique like deep mutational scanning is that we could simply do an experiment where we test all of the mutations on their own and we wouldn’t have to do these sort of complicated population level comparisons to get at the answer. Because when you’re comparing two people in the population, they tend to be different in a lot of ways, and it’s not a very well-controlled comparison. Whereas you can set up something in the lab where you have a gene that does have this mutation and does not have this mutation, and you can really directly see what the effect of that mutation is. Really, people have been doing that sort of experiment for many decades now. What’s new about deep mutational scanning is the idea that you can do that experiment on a lot of mutations all at once.
Starita: And it’s called deep because we try to make every possible spelling mistake. So every possible change in the amino acid sequence or the nucleotide sequence, which is the A, C, Ts and Gs, across the entire gene or the sequence we’re looking at.
Bloom: Let’s say we were to compare me and Lea to figure out why one of us had some disease and other ones didn’t. We could compare our genomes and there’s going to be a lot of differences between them, and we’re not really going to know what difference is responsible. We don’t even really know if it would be a change in their genomes that’s responsible. It could be a change in something about our environment. So the idea behind deep mutational scanning is we would just take one gene. So in the case of Lea, she studies a particular gene that’s related to breast cancer, and we would just make all of the individual changes in that gene and test what they do one by one. And then subsequently if we were to see that a mutation has some effect, if we were to then observe that mutation when we sequenced someone’s genome, we would have some idea of what it does.
Starita: The deep mutational scanning, the deep part is making all possible changes. We have all of that information at hand in an Excel file somewhere in the lab that says that this mutation is likely to cause damage to the function of the protein or the activity of the protein that it encodes. Making all of the possible mutations. That’s where the deep comes from.
TB: How exactly are you doing this? Is it because of advances in computer processing or is it because of a change in approach that has enabled this increase in volume of the different mutations you can look at?
Bloom: I would say that there’s a number of technologies that have improved, but the really key one is the idea that the whole experiment can be done all at once. The traditional, if you were to go back a few decades way of doing an experiment like this, would be take one tube and put, let’s say the normal or un-mutated gene variant in that, and then have another tube which has the mutant that you care about, and have somehow do an experiment on each of those two tubes and that works well.
But you can imagine if you had 10,000 tubes, it might start to become a little bit more difficult. And so the idea is that really the same way that people have gotten very good at sequencing all of these genomes, you can also use to make all of these measurements at once. The idea is you would now put all of different mutants together in the same tube and you would somehow set up the experiment, and this is really the crucial part of the whole thing, set up the experiment such that the cell or the virus or whatever you’re looking at, how well it can grow in that tube depends on the effect of that mutation. And then you can just use the sequencing to read out how the frequencies of all of these mutations have changed. You would see that a good mutation that let’s say helped the cell grow better would be more representative in the tube at the end, and a bad mutation would be less representative in the tube. And by doing this you could in principle group together tens of thousands or even hundreds of thousands or millions of mutations all at once and read it all out in one experiment.
Starita: This has been enabled by that same revolution that has given us the thousand dollar genome. These DNA sequencers that we’re now using, not really to sequence human genomes, but we’re using them as very expensive counting machines. So, we’re identifying the mutation and we’re counting it. That’s basically what we’re using the sequencers for. Instead of sequencing human genomes, we’re using them as a tool to count all of these different pieces of DNA that are in these cells.
TB: At what stage of development is deep mutational scanning?
Starita: It started about 10 years ago. The first couple of papers came out in 2009 and 2010 actually from the Genome Sciences department at University of Washington. Those started with short sequences and very simplified experiments, and we have been working over the years to build mutational scanning into better and more accurate model systems, but that are increasing the complexity of these experiments. And so we’ve gone from almost, “Hey, that’s a cute experiment you guys did,” to doing impactful work that people are using in clinical genetics and things like that.
TB: When you’re at a holiday party and somebody asks you what you do and then they get really into it and they ask you, “Wait, what are the implications of not only personalized medicine but this deep mutational scanning? What’s this going to mean for my life?”
Starita: Right now it hasn’t been systematically used in the clinic, but we’ll get phone calls from UW pathology that says, “Hey, I have a patient that has this variant. We found the sequence variant and this patient has this phenotype. What does this mutation look like in your assay?” And we’re like, “Well, it looks like it’s damaging.” And then they put all of that information together and they can actually go back to that patient and say, “You are at high risk of cancer. We’re going to take medical action.” That has happened multiple times. We’re working right now to try to figure out how to use the information that we are creating. So these maps of the effect of mutations on these very important proteins and how to systematically use them as evidence for or against their pathogenicity. Right now for a decent percentage of these people who are telling them, “Well, you’ve got changes but we don’t know what they do.” We want those tests to be more informative. So you go, you get the test, they say, “That is a bad one. That one’s fine. That mutation is good. That one’s OK. That one, though. That one’s going to cause you problems.” We want more people to have more informative genetic testing because right now in a decent proportion of tests come back with an “I have no idea,” answer.
Bloom: You can also think about mutations that affect resistance to some sort of drug. For many, many types of drugs, these include drugs against viruses, drugs against cancers and so on, the viruses and the cancers can become resistant by giving mutations that allow them to escape from that drug. In many cases there are even multiple drugs out there and you might have options of which drug to administer, but you might not really know which one. Clinicians have sort of built up lore that this drug tends to work more often or you try this one and then you try this other one, but because how well the drug works is probably in general determined by either the genetic mutations in let’s say the cancer or the person or the genetic mutations in the virus or pathogen, if you knew what the effects of those mutations were ahead of time, you could make much more intelligent decisions about which drugs to administer. And there really shouldn’t be a drug that works only 50 percent of the time; you’re probably just not giving it in the right condition 50 perfect of the time. We’d like to be able to pick the right drug for the right condition all the time.
TB: And that’s what precision medicine is about.
TB: Deep mutational scanning as a tool.
Starita: To inform precision medicine.
Bloom: These deep mutational scanning techniques were really developed by people like Jay Shendure and Stan Fields, and Lea and Doug Fowler to look at these questions of precision medicine from the perspective of changes in our human genomes affecting our susceptibility to diseases. I actually work on mutations in a different context, which has mutations in the viruses that infect us and make us sick. These viruses evolve quite rapidly. In the case of flu virus, you’re supposed to get the flu vaccine every year. The reason why you have to get it every year is the virus is always changing and we have to make the vaccine keep up with the virus. The same thing is true with drugs against viruses like flu or HIV. Sometimes the viruses will be resistant, sometimes the drugs will work. These again have to do with the very rapid genetic changes that are happening in the virus. So, we’re trying to use deep mutational scanning to understand how these mutations to these viruses will affect their ability to, let’s say, escape someone’s immunity or escape a drug that might be used to treat that person.
TB: How far along are you on that path?
Bloom: We’re making progress. One of the key things we’ve found is that the same mutation of the virus might have a different impact for different people. So we found using these approaches that the ways that you mutate a virus will allow the virus sometimes to escape from one person’s immunity much better than from another person’s immunity. And so we’re really right now trying to map out the heterogeneity across different people. And hopefully that could be used to understand what makes some people susceptible to a very specific viral strain versus other people.
TB: And so then would your research extend into the mutations in human genes in addition to the changes in the virus?
Bloom: You could imagine eventually wanting to look at all of those combinations together, and we are very interested in this, but the immediate research we’re focusing on right now actually probably is not so much driven by the genetics of the humans. In the case of influenza virus, like I was saying, we found that if there’s a virus that has some particular mutation, it might, let’s say, allow it to escape from your immunity but not allow it to escape from the immunity of me or Lea. That doesn’t seem to be driven as much we think by our genetics, but rather our exposure histories. So in the case of influenza, we’re not born with any immunity to influenza virus. We build up that immunity over the course of our lifetime because we either get infected with flu or we get vaccinated with flu and then our body makes an immune response, which includes antibodies which block the virus. Each of us have our own personal history, not genetic history, but life history of which vaccinations and which infections we’ve gotten. And so, that will shape how our immune response sees the virus. As a result, we think that that doesn’t really have so much of a genetic component as a historical component.
TB: Just going with the flu example, could this result in a future big picture where I go in to get my flu vaccine and it’s different than the one the next person might go in to get?
Bloom: What we would most like to do is use this knowledge to just design a vaccine that works for everybody. So that would just be the same vaccine that everyone could get. But it’s a very interesting … I think at this point I would say it’s almost in the thought experiment stage to think about this. When you think of something like cancer, like Lea was saying, you can use these tools to understand when people have mutations that might make them at risk for a cancer, but that’s actually often a very hard thing to intervene for, right? It’s not so easy to prevent someone from getting cancer even if you know they’re at risk. But obviously if people are able to do that, they’re interested in spending a lot of money to do it, because cancer is a very severe thing and you often have a very long window to treat it.
Something like a flu virus is very much at the other end. If I had the omniscient capability to tell you that three days from now you’re going to get infected with flu and you’re going to get really sick, we could prevent that. We have the technology basically right now to prevent that, if it’s nothing else than just telling you to put on a bunch of Purell and don’t leave your bedroom. But there’s also actually some pretty good interventions including prophylactics to flu that work quite well. But the key thing is, right now we think of everyone in the world as being at risk all the time and you can’t be treating everybody in the world all the time against flu. There’s just too many people and the risk that any person …
Starita: Not that much Tamiflu on the market.
Bloom: Not that much, and the risk of it … So I think to the extent that we could really identify who’s at the most risk in any given year, that might allow us to use these interventions in a more targeted way. That’s the idea.
TB: And how does deep mutational scanning lead to that potentially?
Bloom: Yeah. So the idea, and at this point, this is really in the research phase, but the idea is if we could identify that say certain people or certain segments of the population, that because of the way their immunity, let’s say, is working makes them very susceptible to the viral mutant that happens to have arisen in this particular year, we could then somehow either suggest that they’re more at risk or, as you suggested, design a vaccine that’s specifically tailored to work for them. So that’s the idea. I should make clear that that is not anywhere close to anybody even thinking of putting it into economic practice at this point because even the concepts behind it are really quite new. But I do think that there’s a lot of potential if we think of these infectious diseases not so much as an act of God, where you just happened to … someone sneezed on you as you’re walking down the street, but actually a complex interaction between the mutations in the virus and your own either genetics or immune system, we can start to identify who might be more at risk for certain things in certain years, and that would at least open the door to using a lot of interventions we already have.
Starita: The first year was three years ago, and some very enthusiastic graduate students started it. Basically, it was almost like a giant lab meeting where everybody who is interested in this field came. Somebody tweeted it out and then all of a sudden people from UCSF were there and we’re like, “What the heck?” It was great and we all talked about the technology and how we were using it. The next year, the Brotman Baty Institute came in and we’re like, “OK, well, maybe if we use some of this gift to support this, we can have a bigger meeting.” And then it was 200 people in a big auditorium and that was great. And now this year, it’s a two-day symposium and workshop, and it’s also co-sponsored by a grant from the National Human Genome Research Institute. But now we’ve got hundreds of people, so about 200 people again, but now flying in from all over the world. We’ve got invited speakers, and the workshop, which is Tuesday, is a more practical, “If you’re interested in this, how do you actually do these experiments?”
TB: What’s driving the interest in deep mutational scanning?
Bloom: We are starting to have so much genetic information about really everything. It used to be, going back a couple of decades, a big deal to determine even the sequence of a single flu virus. It was totally unthinkable to determine the sequence of a human genome, right? If you don’t know what mutations are there, you don’t really care that much what they do. Now we can determine the sequence of tens of thousands of flu viruses. I mean, this is happening all the time, and we can determine the sequence of thousands, even tens of thousands of human genomes. So now it becomes, as Lea said, really important to go from just getting these sequences to understanding what the mutations that you observe in these sequences actually will mean for human health.
See this site for more on the Brotman Baty Institute for Precision Medicine and the Deep Mutational Scanning Symposium and Workshop, Jan. 13 and 14 in Seattle. The symposium is free to attend if you’re in the Seattle area, and it will also be livestreamed, with archived video available afterward.