What if you could sequence a tumor? And then replicate that tumor? And what could you do if you applied AI to in vitro fertilization? These three stories and more showcase the future of personalized medicine, with one exceptional professor at the heart of them all: Dr. Olivier Elemento. Dr. Elemento is Professor of Physiology and Biophysics at Cornell Medical College, a Walter B. Wriston Research Scholar, Professor of Computational Genomics in Computational Biomedicine and Associate Director of the Institute for Computational Biomedicine, Director of the Englander Institute for Precision Medicine and Associate Director of the Institute for Computational Biomedicine. That’s a lot of titles!
Bobby: Welcome to Loka's Podcast, What Fascinates You?—conversations with entrepreneurs, engineers and visionaries who are driven to bring innovations to life. I'm your host, Bobby Mukherjee. Today I want to dive into three stories that showcase the future of personalized medicine with one exceptional professor at the middle of the mall, Dr. Olivier Elemento.
Dr. Elemento is the Director of the England Institute for Precision Medicine and holds several positions at the Meyer Cancer Center at Cornell Medical College, including associate director of the Institute for Computational Medicine, and most recently Associate Program Director of the Clinical and Translation Science Center.
Dr. Elemento's Research Group combines big data and AI with experiment. And genomic profiling to accelerate the discovery of cancer cures. They have been published in over 200 scientific papers and featured in major media outlets, including Popular Science, Gizmodo and The Huffington Post. But today we want to explore just three of these stories: What could you do if you could sequence a tumor? What could you do if you could then replicate a tumor? And what could you do if you applied AI to in IVF? Okay.
So for the soundcheck purposes, in your copious free time, are you more of a book reader or do you like watching Netflix or HBO? How do you decompress?
Dr. Elemento: That's a great question. I read a lot of books actually. I, actually I don't read books. I listen to books I've discovered, actually unfortunately I should have audio. But it so many books are available now in audio, and Audible or even the New York Public Library actually has a million books that are available for free, and they're just amazing. And I listen to a lot of audio books these days, which I realize is awesome because you can go to the gym and listen to a book. You can, you do your shopping and listen to a book. You can do the laundry. And actually it's funny because actually now I want to do the dishes, I wanna to exercise more because it just gives me more time to listen to books. It's actually changed my life in a positive way.
Bobby: No, that's great. I think we, if you don't mind, I think we use exce excerpts of that. That's like a great joke to put in the middle somewhere.
Bobby: I look forward to doing the dishes. I think it's not the quote people are expecting when they listen to Dr. Elemento podcast. So tell us a little bit about your background and how you came to be in that lab that you're talking to us from.
Dr. Elemento: My background originally is in mechanical engineering. That was a very long time ago. I've always had a fascination regarding engineering and how to build things. I had actually a fascination regarding robotics, from from an early early age. But I always maintained this really great interest in both two really important topics that are now important in my life, which is biology and informatics. When I was doing my studies in mechanical engineering. I really maintained the student great interest in informatics. In fact, into many courses in informatics, in computer science. When I was in college. Towards the end of my engineering studies, I was interested in maybe expanding, the scope of my studies if you want.
And so I decided to do a a master's in the artificial intelligence. And I had a great time doing this, a master of learning new things. It was, artificial intelligence many years ago long before, it was cool. Be walking in ai. It was essentially during the AI winter that I think sort did this curriculum.
But I also, as I say, kept this interest in biology. Towards the end of my studies in artificial intelligence, I got a chance to work with a great form mentor mine on the topic that involved analysis of biological data. And it was using self-organized maps, basically a special type of neural network that used to be very popular back then.
And the sort of confluence of these different aspects led me to be very interested in computational biology and how essentially to apply AI to biological problems. I started doing a PhD in computational biology. Had a lot of fun doing that. It was more a theoretical course, so the type of PhD where I was trying to reconstruct histories of gene duplication and.
And the genome has a lot of copies of the same gene, slightly different copies, and I was trying to reconstruct essentially a tree of the evolution of these copies. How do you start with one gene and you end up in the human genome, for example, with a 50 genes each other visible different. So that was kinda a PhD, but it really gave me a great passion for analysis of genomes, which I continue to do during my postdoc Princeton in.
I learned a lot more about genomics, learned a lot about how we can understand what genes are express and where in the body, using information about the DNA to understanding the switches and the DNA that turned on and off genes. And that led to postdoc to take on a faculty position at Cornell where basically decided to apply.
All of this background in genomics to disease that I got really interested in was cancer and applied a lot of these thinking and analysis to cancer, genomics to the understanding of cancer cells. The understanding of what mutations drive cancer, why is it that people get cancer? What is the sequence of event that gives rise to a cancer? And we were reasonably successful in un understanding, all these aspects. And that led to a lot of different ideas, including what I'm doing now, which is. I'm the director of an institute that's called the Institute for Precision Medicine that's essentially trying to apply all this concept that we learned about the genomics of cancer and other disease to personalize medicine, to try to make medicine as personalized as possible, which is also enabled by revolutions in technology, which allow us, for example, to sequence somebody's genome.
In just a few hours, we can actually detect, analyze the blueprint of disease in a very short amount of time, and now analyze how to understand those genomes and how to treat patients in a way that's connected and matched to the genome.
Bobby: That’s a quite a fascinating journey, and it seems curiosity was your guide. As you dug deeper and deeper, you found these launch points that took you to where you're today. I'm just curious, like in an alternate universe when you made a decision many years, To start off by studying mechanical engineering. When you started that journey, did you have a slightly different destination in mind at that moment in time? What did you think you were gonna do when you grew up?
Dr. Elemento: That's a great question actually. I was I studied mechanical engineering because had a initial passion for robotics. I was always very intrigued by the potential of robots to, achieve things that are to do for. Including, for example, sending robots to different planets, to the moon or to mo, I always thought that was just an amazing feed of engineering to be able to, create robots that can do, smart things and enable things that are completely possible by humans.
So this, very difficult by humans. And actually I do think that, the potential of this is still, just incredible. I think in the, in my field, in, in medicine, there's so much that's going to be automated. There's so much where I think robotics eventually will be able to help automate things manually, even doing experiments.
A lot of what we do in our institute is manual experiments. We're essentially doing experiments that take a lot of labor and are very difficult. I do think a lot of these experimentation is gonna be automated, is gonna use robotics. So I'm hoping at some point in the future I'll be able to look back in robotics into things that we do, more so than we do now to be able to automate, certain things that I think can be automated.
Bobby: One thing that I've always really loved about. A university setting is that it creates a natural environment for deep specialists in very different domains to have natural proximity to each other, to collaborate. And I'm just wondering if you've seen any of that come to life in your work at your lab?
Dr. Elemento: Absolutely. I can tell you, what we do is fundamentally multidisciplinary, analyzing somebody's genome, involves all kinds of different aspects, collection of samples physically, surgery right in the first place. Or bio doing biopsies, processing the samples to be able to extract DNA from samples, which can be, sometimes difficult.
We do a lot of single cells sort of type analysis. We have to take a piece of tissue and essentially extract all of the individual cells in the tissues so that you can process from one set at a time. And this processing samples in this way also takes a lot of e. The technologies that we use for sequencing, for analysis of single cells are also just, amazing of engineering, just incredible technologies that we have access to.
And we need people who understand these technologies to be able to get optimal results from the technologies. And then, there's computer science and computational biology and ai. We need people like this as well as part of a team to be able to understand the gigantic amount of data that we generate on each patient.
But also contextualize what we see in one patients in the universe of all of the other patients that we've seen that, yeah. As a lot of what we do in precision medicine is kinda a patient like mind type of analysis where, we try to position one patient based on the data that we collect in the universe of other patients.
And if something worked in another patient that's similar to yours, you can infer that, treatment may walk in, in the patients with similar feature or similar mutations, for example. We have to be essentially able to be comfortable with this kinda analytics involving large data sets and technologies that are very.
Everything goes back to the physicians. So we have to also be comfortable with interactions and communications and collaborations with physicians. Cause at the end of the day, they treat the patients. We don't. So all of the information that we extract from genomes, we need to be able to communicate it, to put it in the report in a way that's succinct and clear so that physicians that we work with will be able to consume this information and use it to treat the patient.
Bobby: Sounds like a really rich collaboration. So in your experience with working with, these different constituents, so I'm just curious, have you noticed a mindset difference between, say, physicians and, data scientists, for example, do they approach a problem in a very different way that you've noticed?
Dr. Elemento: I think people do tend to approach problems in different ways, at least at the very beginning. But I think what I, we are trying to promote here is really this notion of community where, people really have to interact with each other to be able to solve this complex problems.
And as we interact with each other more and more, I think we start more and more to speak the same language and start being able to understand each other, in a much more effective way. And I'm seeing that quite a bit. When we. This work a few years ago, there was a lot of initial kind of silos if you want, people doing their own things, by having people communicate with each other, meet with each other, and finding as many forums and opportunities to enable this communication.
We are the point now where, it feels like one community. It's everybody's, talking to each other. There's a lot of interactions between people who come from different fields and they understand each other much, much more than they used to. It's actually very rewarding from my point of view to see that there's a lot of cross-disciplinary interactions that are actually quite smooth and very productive.
And I think this is what this field is really about. Multi disability interactions is really, critical to this field. And honestly, I think it's just critical to many field. I think discovery has come from, taking, different people, different mindsets and having the ideas where ideas clash and merge.
And I think this is, where fun science is, is happening, when when you have, interactions of ideas and people who don't necessarily always, nobody talk with each other or interact with each other, that's where the magic can happen.
Bobby: Let's dig in a little deeper into one of those areas. You were telling me about how you used AI to treat uterine cancer. So I was wondering if you could give us a bit of a backstory of how did that project begin?
Dr. Elemento: Yeah. The context here is a kind of a systematic attempt at Cornell to detect and assemble the blueprint of disease. Especially cancer.
For every cancer with advanced disease, we sequence basically the DNA of those patients and we try to understand what's happening. We try to understand, what are the sequence of mutations that happened in this one patient and what can we do about these mutations? How can we use this very detailed understanding of essentially the map of somebody to be able to treat the patients in a very effective way. And I think when we do this, we just realize that, in every cancer patient that we sequence, there's a lot of opportunities to treat them because a lot of mutations basically changes in DNA that make the tumor cells different from normal cells.
That's what we call mutations. A lot of these mutations are. What we call actionable in a sense, and often there is a drug that can target the product of mutations. What's great is that the field of cancer therapeutics, has also grown quite a bit in the past few years. We actually have now many drugs that are available.
That can target a product of mutations that we see in patients. Part of the challenge is to match them. It's you sequence somebody's genome, you find a lot of, and you have a lot of different drugs that are available. How do you match, how do you find among the list of drugs that are available, the best drug to treat each individual patient?
So this process involves a lot of work, not only to analyze and sequence the genomes in the first. But to interpret genomes, genome interpretation is really what's hard. It is where AI is really becoming quite fundamental because. Despite the fact that, there are some mutations that we see that are low hanging fruits, easy to find, easy to detect.
You see them, where they're important. The genome is very complex and the genome typically of a cancer patient has potentially hundreds of thousands of mutations. Many of them, we don't understand what they do. AI is being used now by us and others in the field as a way to essentially better read the genome and better understand the genome.
And so many mutations that are interesting, emerge as better understood as a result of this AI analysis. And so the more we understand the genome, the more we can match the genome and what we see in the genome. To a list of drugs that we have access to and our physicians have access to. And and that's really the broader context is really this very detailed building of the map of genome that requires ai.
So in the context of this, we've seen a lot of really interesting things. We we've seen a lot of mutations that we did not expect to see in some people with with cancer. You mentioned the case of this patient with mutation cancer. We saw in that patient a mutation that we almost never see in mutation cancer, a mutation that we typically see in other types of cancers.
In taking this particular case, it was a breast cancer mutation, a mutation that you see often in, in breast cancer, but almost never in mutation cancer. But we know now that the right way to think about cancer is not to only think about where the cancer is located in the. The right way to think about cancer is actually to understand what are the mutations that are found within each individual cancer.
That gives us a lot of really useful information. In this particular patient, we were able to treat that patient with a drug, but nobody is used in breast cancer because breast cancer, patients have more often these mutations. And were able to help these patients by having this drug be quite effective in this particular case.
And it was only due to the fact that this mutation was one patient and was clearly driving the cancer and therefore was sensitive to inhibition as a result of treatment of the drug.
Bobby: Yeah, that's incredibly powerful. So let's take a step back and try to understand what building blocks became available. More recently that maybe were not available five or 10 years ago that allowed for this jigsaw puzzle to come together. Maybe for some of our listeners, you could explain. Some of the advancements that have happened in just even basic sequencing that was not possible five or 10 years ago.
Dr. Elemento: What's really been amazing in the past few years is ability to sequence genomes from beginning to end in just a few hours. We can sequence somebody's genome now in 24 hours, potentially in less use. Some of the very recent sequencing techniques that have emerged. You can actually do that maybe in five. It is costly to do that. Essentially the more, the quicker you get, you wanna get your genome, the more, the higher the cost is.
But again, when it comes to, treating patients, I think, very often I think it's so important that it's spend additional, extra. On sequencing. I think that's something that was not available even, five or 10 years ago. This ability to sequence somebody's genome, so quickly, and it's an incredible achievement to be able to have, such broad access to these technologies.
This technologies are just amazing. They don't cease to amaze me. The way we sequence the genome now is really, it's really cool. Essentially what we do is we take all of the DNA from, let's say a lot of DNA and we chop the DNA into tiny fragments that we essentially put in a machine under sort of glass slide.
We attach the fragments on the glass slide. Essentially using a process of l sequencing, we sequence millions and millions of these DNA fragments. At the same time, we actually turned the process of sequencing genomes into an image analysis problem. So basically we do sequencing by synthesis. We essentially reconstruct the DNA of each fragment cause of the complement of DNA. We have two strands and we know stuff, and one strand reconstruct the other one. We add nucleotides one at a time with a specific label of fluid label. And essentially we recall the attachment of individual nucleotide as essentially a new sequencing event that allows us to visualize the way that nucleotides are added one at a time, and essentially sequence millions and millions of these fragments at the same time. The challenge becomes a computational challenge where you have these millions and millions of fragments. You have to essentially put them back into single genome.
It's kind piecing the puzzle back together if you want, and you have to look at the overlap between fragments and reassemble genomes from scratch. That actually requires a ton of computation to do well. And that's actually quite hard in the context of cancer genomes, which have a lot of mutations that can be very difficult to reconstruct and very complex.
In addition to this, you have a lot of improvement recently in the field in terms of single cell analysis. I was telling you. We realize that, in, if you look at a cancer, what we've done in the past is to essentially look at the, the average DNA of the cancer.
But in fact, cancer has billions and billions of cells. What we should try to be doing is to see, is to analyze the. Each one of the cells one at a time, because we know that there's gonna be a lot of differences between cells in the cancer, and we need to be able to capture this heterogeneity.
And so new techniques and sequencing now allow us to sequence all of individual cells in the cancer and understand the viability of each of the cells. This is not something that yet had applicable in the clinic. Something that we want to put in the clinic at some point, but that's the future of sequencing, is to move away from this kind of average DNA of a tumor to analysis of each individual cell to understand the huge number of possible mutations that even, one cell can have in a tumor. That's actually very important because we treat cancer patients with targeted drugs that are very focused on targeting the mutation in the particular gene. What that means in practice is that if you have lots of cells and they make random mutations here and there, it's not impossible that one cell is gonna have a mutation, a random mutation that makes the drug unbind to the target, and it just only takes one cell. If there's one cell that has to be mutation, the cell is gonna be resistant to the drug and it's gonna keep growing and it's gonna basically take over the whole population. And that's how you end up with a cancer re resistant to treatment. It's because of this constant evolution of cancer and the fact that it doesn't take a lot of mutations to have a cancer that's resistant to treat.
Bobby: The ingredients then that led to the success. You have the speed at which you could sequence and the fidelity of that data. Then all the sort of computational improvements that have allowed the analysis of it. It's interesting to me that another ingredient though was the fact that there was this breast cancer drug already approved by the F D A and out there. I don't know, we could go back and check, but I'm guessing that drug had actually been out for a number of years.
Dr. Elemento: Absolutely. It's actually one of the first targeted drugs to be invented. It is actually the first targeted drugs to be invented. It is actually a very interesting drug. As a drug that's been immensely successful. It targets the product of a gene called her two, and the drug is actually an antibody, and the drug is actually essentially a protein that's able to recognize a partially mutated protein and bind it and basically block it. The fact of blocking the action of that protein, such as destabilizing the cancer cells and leading the cancer cells to commit suicide to apoptosis. So it's actually an amazing drug and it's been a lot of drugs that have come out since, use a similar sort of type.
But this is one of the many great drugs that have been invented. And there's just so many, the field of drug discovery now is just amazing. Cause, there's so many modalities that people use from, antibodies to small molecules to therapeutics, as where you can essentially re-express particular gene in a cell to achieve a particular result There's so much that exists now. There's so much happening in the field of drug discovery, and so the list of drugs that we gonna, continue to have is gonna keep increasing, which is great because then you can even better match somebody's genome to this of what we do is to learn better how to drugs.
We actually not very good now at assessing how to combine drugs to treat patients. It's actually a complex problem, how to find combinations of drugs that can maybe kill cancer cells, for example, better. And it's actually a pretty complex problem because, there are so many drugs and choosing the cell, the sub subset of drugs that are effective in a particular context is difficult.
I do think this is an amazing opportunity for AI and for machine learning and for data science to help. I always say, there's so many drugs that it's gonna be impossible to pipe fast enough to test every combination. That's not gonna happen. So we need to come up with a, maybe a conceptual framework or some kind of a, I dunno, virtual model of cancer cells or virtual model of disease.
Like a way to simulate what would happen if you treat a cancer with, lots of different drugs because we're not gonna be able to experiment to do experiments quickly enough to actually test all these different combinations. So it's a new mindset that I think the field needs to adopt. And I think, as I say, machine learning and AI and data science, I think is gonna play a big role.
Bobby: So it sounds like a mindset that is related to another topic I wanted to talk to you about, which was this idea you were telling me about a situation where you are making copies of a patient's tumor. Small copies and then, I think I saw footage of a robotic arm trying different drug recipes on these different tumors to find one that's effective.
And it seems like you returned back to your mechanical engineering robotics roots. Correct. With that idea. Tell us more about what that is.
Dr. Elemento: Yeah, this is I think also a an amazing opportunity for patients. We sequence genomes over time and as much as, as we were discussing, there's so many opportunities now to use the genome as a way to treat patients.
We also pretty frustrated because the reality is that genome, there's a lot of mutations that we don't understand, that we don't understand the function of, and there's a lot of mutations for which ways not work. So we decided to take maybe a distinct approach, Essentially hedge our beds.
If you want to try to figure out what else could potentially be effective in patient. What kind of drugs could potentially be effective in terms of killing cancer cells without having to treat the patients or without having to do clinical trials on the patients first. These are idea approved drugs they have already been shown to be effective and safe.
But again, I think the challenge is always to find the ones that are effective on a particular patient. There's a great approach now that you can take, which is to take cells from the patients and then grow them outside the. And then essentially grow them to the point where you can make enough cells to create mini two moles and the mini two moles, because they are made of the cells from the patient, from the tumors are from the patient, and are essentially copies of the patient's tumor. And we can make a thousand, or maybe even 2,000, 3,000, maybe 5,000 full cells that grow quickly. Such mini copies. So once you have, let's say, a thousand copies of a patient's tumor, what you can do is to try a different drug. It's like an empirical testing of drugs essentially, as opposed to guessing what drugs may work based on the genome, you actually test directly under a patient's cells.
What are the drugs that are actually effective at killing tumor cells from this particular patient? We can't do that manually. We have to use a robot to do this because we have to test a lot drugs and combinations and then, we have to measure different time points. There's. Repetitive work and too much work to do this manually, but it turns out to be, it turns out to be a very effective approach for identifying drugs that work on individual patients.
We basically, just don't have to guess. We see what works and what doesn't. And this is something that is not yet a clinical has say, available for patients, but we hope very soon, maybe one year or two years from now will be something that we can offer to. The beauty of it is that it also allows us to make discoveries, right?
Because as we start doing this, now for research, you know that we'll do this more and more moving forward. We can go back to the genome and correlate the response of the different drugs to the mutations that we see in the, in, in the cancer cells. And the idea is that if we always see that a mutation is connected to respond to a particular drug, then you know, at some point we may actually not need to do the drug testing.
We can maybe just look at the genome and find this mutation and say, look, we know that every patient with that mutation responds to this particular. So we are actually learning also how to better understand the genome based on this very large scale analysis and drug testing analysis that we do for for many patients now.
It's a very exciting way to also find, new treatment options for patients and and we're learning about cancer cells and genomes cells also in the same.
Bobby: I read something you had been quoted as saying, which was fascinating to me, which is that you have distinct age related differences between tumors. And so you almost have to treat them as different patients and different cases, which I thought was fascinating. And so could you maybe expand on that a little bit?
Dr. Elemento: Yeah. This is research and we published a few months ago now where we looked at a very large number of cancer patients and genomic profiles from cancer patients, basically DNA profiles.
And we compared patients whose age is, on higher versus patients versus younger. And we ask ourselves, do we see a difference in terms of the DNA sort of makeup of younger patients with younger, the, from younger patients versus older patients? And as it turns out, we do see lot differences.
And it does match what we've, what has been seen clinically in, in, for many tumors, which is, it is often much more difficult, for example, to. Successfully treat patients who tend to be older and nobody really knew why. I think, you can make all kinds of assumptions about, the chemotherapy is, gonna have more side effects on older patients, but what we see now is that the biology of the cancer is different when the patient is old versus when the patient is younger. Now, the age of a patient is impacting the biology of a cancer in the first. So patients essentially get different cancers or different types of cancers when they're older or when they're younger, even though the cancer may be called the, it's a colon cancer.
But actually, if you look at the map of a disease, the blueprint of a disease that actually turns out to be different between younger and older patients. What we are saying now is that it does matter to look at the age of the patients and we should essentially maybe potentially, Younger and older patients in slightly different ways to reflect the fact that the biology disease is different.
But honestly, this is really the fundamental idea of our institute and what we do at Cornell is really to understand what makes every patient unique as we have a technologies to be able to do so now. All the technologies that I mentioned from single cell analysis to DNA sequencing, we do see that every patient is unique.
When two patients are diagnosed with breast cancer, if you actually look at the tumor, if you look at the detailed analysis of the tumor, you actually see that, despite the fact that the cancer is a missing location. The disease is different. The disease is caused by different mutations.
It has, maybe different immune cells connected to the disease. It has a lot of there's a lot of uniqueness in the disease and that I think, uniqueness, I think has been a bit discounted in the past. And I think we're trying to reveal, what makes every patient unique so that we can treat each patients in a very unique way according to what we understand about the disease of each individual patient.
Bobby: Yeah. So that to me is just taking precision medicine, personalized medicine to the next level.
Dr. Elemento: And that's exactly what it is. Precision medicine is taking personalized medicine to the next level. Medicine has always been personalized, right? There's no such fingers non personalized medicine that's in part of mine, but with technologies such output sequencing or single cell analysis or robotic screening for drugs, for these many copies of tumors, we can really take personalized medicine to the next level to make it very personalized, based on the understanding of the disease as we discussed.
Bobby: Staying on that precision medicine front, let's talk about another area which is around using AI for I V F. So from friends that have gone through the process, the IVF journey can be really emotionally draining. And to me, as a tourist on the outside looking in, it just seems. Incredibly trial and error filled with really rough emotional highs and lows for the couple that's trying to have a baby. And tell us a little bit about how AI might be playing a role in I V F and the work you're doing there.
Dr. Elemento: The context is really the same, right? I do think that all these technologies that we've discussed can be applied to all kinds of different disease. We talked about AI for cancer. I think what we've seen in the past few years is that you can use these technologies for all kinds of different applications.
I think one great use of AI medicine is for image analysis where, you can essentially automate a lot of the, analysis of images that used to be done manually or visually by, one trained. You can automate it to a point where, you can really help physicians be very productive by automating a lot of the processing of images, analyzing, not one image, you know at a time, but like a1 thousand images, at a time, which is, also easy to produce.
We became interested in the idea of applying some of these AI techniques to embryology into IVF, a few years ago, some colleagues, Because we realized that the embryology lab in an IVF clinic is producing a tremendous amount of image data on the embryos that we look at after fertilization.
So in the IVF process, there's fertilization that happens typically using a process called where injection of sperm into an. And then what embryologists do after that is essentially they watch the embryos. As we developed, as, initially two cells and then four cells and the late cells and so on. They look at different time points and they assess the quality of the embryos based on what they see. Embryo, small things. We have to use a microscope to be able to visual. Actually now essentially we call movies of the development of these embryos. And we realized a few years ago as well that the process of analyzing the quality of these embryos is actually not necessarily as robust as we thought it would be. In a sense there's a lot of differences between individuals who, watch the same movie, but may actually come to a different conclusion. When comes embryo's important. The quality embryo determines whether embryo implanted. That is really a critical step of the process.
So we got access to a very large number of movies of such embryos after fertilization. We had a lot of information about these embryos, whether, for example, the embryo after implantation would give rise to your life, whether the embryo would implant, whether there was abnormalities of his embryos after genetic testing.
And we had a kind of a the crazy idea to apply AI to this gigantic database of videos and images that we have. And that really, was amazing because we saw that we would be able to get results in terms of predicting quality of embryo or chromosomes. Actually, that is actually comparable, if not better than training and also does not have the kind of viability that you see between embryologists. And I think really it gave us to the idea that, we can apply AI to IVF and we're doing that, we're doing this for other types of problems. But I think, this is a great way, I think, We think to eventually maybe democratize the process of IVF, which is still re restricted now to a few clinics.
It's not something that is fully accessible. It's expensive but we think that using AI you can only make the process of IVF a bit more standardized, reproducible if you want, but also make it cheaper by essentially having sort of computers do some of the work that is done by humans and essentially allow embryologists to be quite productive with what they do as opposed to spending a lot of time watching embryos and looking for something, small. Like why not have a computer just do the work for you? And then, you look at where the computer produced and then you say, okay, that makes sense. That the computer is saying the right thing. And therefore, this is how patient this is what we need to do next for patients.
Bobby: So this would be yet another example where it's not the case that, decades and decades from now, this is gonna make the need for embryologists redundant. It's more embryologists plus AI will be much more successful than just embryologists by themselves.
Dr. Elemento: Absolutely. I think that's really critical. Embryo obviously do a lot of the process, including, for example, the fertilization. In fact, you can make a case that, the real value of an embryologist is really this, very difficult process of fertilization because you can damage the embryo.
There's lot of things that can go wrong. In the process, if you can allow them to do more of that and spend more time and, being as careful as possible and picking the right sperm and so on, you think that it's actually beneficial for everyone, for them to do that. So if you can reduce the time they looking at the embryos develop, looking for the law, and actually help them be more productive. And that's really what we. AI needs to help people, physicians, and people in the medical field be as productive as possible. It should not replace human beings. Cause at the end of the day, obviously as a patient, you want have this interaction with another human regarding your care, and you want actually get messages that relate to you in ways that you can understand.
You want to have all of the the empathy and the warmth and the communication with a doctor or somebody in the medical field that's really critical. This is actually a big component to medicine. This, sort of interaction between human. We don't want, we don't wanna take that away.
We actually want to give more of that. We want to actually automate everything that can be automated so that, physicians can spend more time with their. Because I think that's really the part where they add tremendous value to the process, to the medical process by applying these AI techniques to IVF.
Bobby: The hope and the outcome is that the success rate of the procedure is higher, is the driving force behind that. The choice of which embryo is better, or like what are the key components that lead to the better outcome?
Dr. Elemento: So the choice of the embryo is really important, especially for couples that have multiple embryos. Choosing the right one is really critical. If you choose the right one really quickly, you can, for example, avoid, maybe doing another cycle, right? You can, more quickly be successful. And that's really important, for doing ivf, everybody's aging, and the older you're in the IVF process, The harder to have a good outcome.
So if we can facilitate a good outcome as quickly as possible in the first cycle, I think that actually helps. You know tremendously. But I think the other thing that's happening, INF, is very, now we're not very good at finding right highest quality embryos. What is often done is that you, that physicians will implant multiple embryos at the same implant, let's say two or.
And the problem with this is that in the goal here is you kind edge your bets, to hope that one of 'em is gonna, be be successful in terms of implantation. The problem is that very often what it gives rise to is complications because, often it rise to.
Multiple pregnancies at the same time. And it may sound good, to have, multiple babies at the same time. But the problem is that very often it comes with complications. Picking the right embryo, and implanting that one to maximize success. Embryo is decrease complications that okay, sometimes in IVF with multiple pregnancies. So it's a combination of that. We are applying AI to all problems so that they can decide when to conceive at a later phase.
But right now there's not a lot of science when it comes to selecting which embryos freeze and think that the same techniques that we use selection can also be used, can maybe a smaller side of, so that, we can also reduce the cost and maybe also improve the outcomes down the line.
Bobby: I love the spirit of this. Trying to democratize IVF because it is so expensive for a lot of people.
Dr. Elemento: It is very expensive, and as populations are aging very quickly. I think, many countries actually having, pretty major fertility problem. And I think that IVF is, not the only answer, but I think it's going to be an answer to some of these issues.
Democratizing IVF is going to be very important. I think a lot of countries are actually in the process of providing better coverage for fertility treatment. And, and I think if you can make the process more streamlined, more effective decrease the cost, I think it's just gonna help, it's gonna help make the process more available to more people. It's just gonna help a lot of people.
Bobby: I can only assume that finding talented folks to bring onto your team is a struggle because those people, I'm sure are in high demand and you're looking for very specific things. What do you look for in candidates when you're bringing 'em into the Elemental Lab?
Dr. Elemento: The challenge that we have is that we are in need of people who can understand both computers and programming and statistics, and also the biology in the medicine. This is a real challenge that we have. We're looking for, a rare breeder people who can really, do a lot of things, and, and that's very hard, thankfully, because, we are trying to address important problems.
We have some level of visibility and that you helps us in terms of being attractive to two people. But, that's the kind of profile that we look for. People who are comfortable in multiple fields who have the ability to be comfortable in multiple fields. I do tend to, And it's hard to be good at everything.
So it's almost hard to have somebody, who's like perfect, amazing at, everything that we work on. So that's why, we tend to maybe sometimes hire specialists in one thing, and we teach from, let's say, what, what they're liking, whether that's biology or or computer science.
But it is important really to be flexible and to, be willing and eager to learn new things in, in, in our field because, Know, it's really an interdisciplinary field by definition. And it's hard for people who come into a field to stay in their little boxes.
You have to be flexible and be eager to learn and be, willing to to come out of your shell if you want to really, learn new things. I think it's so that's the kind of profile that ver we're looking for, these days. But, obviously we also, need people with ai background.
We need people with computer science background. We need we need people with those back. One thing that we have for us being in the university is that we train a lot of these people. We have PhD programs where, we train people to acquire those skills, to do the research that we, and we try to as use pipeline, we train of them when they're trained.
So that's, I think, something that's unique about, maybe in some ways university is that we have also this training sort of program that's kinda built. Our business model if you want, and it's a great source of talent as well.
Bobby: These are really meaningful problems that you've been chasing solutions for. I'm certainly inspired. I know it's gonna be the same for a lot of our listeners, those of us that wanna dig in further, learn more, maybe even participate in some way. Do you have any upcoming events that you'd like to mention or interesting papers or websites or articles that you could send folks to to do some homework afterwards?
Dr. Elemento: I'll be happy to send the the website of our institute. We have a lot of material that can be potentially used, to learn more about things that we do. As I said, we have a pretty broad. So the spectrum of projects, a lot of really exciting. It's just an amazing time now to be in medicine, biology.
These technologies I was telling you about, just changing everything about how we understand disease. It allows, this really incredible understanding of disease, down to resolution that we just, could only dream of. A few years ago, and attached to this we have was happening in the biotech world with all these new modalities, as I said, mRNA, therapeutics, antibodies, and early detection disease, from liquid biopsies.
There's so much going on now. It's such an exciting field and I'm so excited about the future of this field. I think it's just incredible to be in this field right now. So much going on.
Bobby: I definitely feel that way. I am, again, so grateful for you making the time. I know our audience is gonna get time out of this and Dr. Elemento, I'd just like to thank you so much.
Dr. Elemento: Thank you having me, it's been a blast.
Bobby: That was Dr. Olivia Elemento; scientist, professor, cancer fighter, and a driving force for AI and personalized medicine. If you're interested in learning more about Dr. Elemento and his research, go ahead and Google Elemento Lab and if you enjoyed our show, please like and rate us. Until next time, this is Bobby Mukherjee.