In this episode of Qubites, we welcome Shahar Keinan, and we will explore the use of quantum computing in drug discovery. We will delve into the ways in which quantum computing can accelerate the process of discovering new drugs and help in identifying new potential drug targets. We will also discuss the current state of quantum computing in this field and its future potential.

Transcript

Rene – Hi! Welcome to QuBites, your bite-sized pieces of quantum Computing. My name is Rene from Valorem Reply and today we're going to talk about Quantum Computing for drug Discovery. Very important topic, and for this I'm honored to have a special expert guest today, Shahar Keinan. Hi Shahar and welcome to the show, how are you today?

Shahar - Very good, thanks! How are you doing, Rene?

Rene - I'm doing fantastic. Can you tell us a bit about yourself and your background as it relates to chemistry and Quantum and all of the goodies?

Shahar – Sure. So I'm a computational chemist by training. I did my PhD from the Hebrew University of Jerusalem, came to the US for a post-doc and stayed. I started Polaris Quantum Biotech with my co-founder Bill Shipman who is Polaris CTO and myself about three years ago and we are located in Durham, North Carolina and we're a virtual company. So we have people all over the U.S.

Rene - I love this modern work style! Alrighty, so let's dive into our today's topic. First of all, let's make everyone basically on the same page. So tell us a little bit what is the process of drug discovery like and what is so special about you know these more bond personalized precision medicines?

Shahar – Drug discovery is a very lengthy and complex process. The part that we are interested in is in the early stages of drug discovery, sometimes also called drug design. After you know what the disease is, and understand the biology of the disease, you identify a specific protein which is part of the cell machinery that is not working correctly or is overworking. You then try to either stop that activity or enhance it. This is when you're looking for a small molecule that will become your drug later on. We come into the process at this point. We are the group that can find those small molecules faster and more efficiently. When we find these molecules, they go through more testing, both against this specific protein and then against other proteins to make sure that there are no side effects in all sort of models. After all of that is finished, then those small molecules go and become somebody else's problem and then they start doing clinical trials. These trials can take five to seven years before these become a drug that you can sell in the market. So, what we are interested in is really the early stages of drug discovery and drug design. We are interested in that because the way people are doing it today is through steps of trial and error. You design some molecules, you test them, you see if they're good or not good, you learn something, and then you go to the next step. Each one of those steps looks at about a thousand molecules and it takes the whole process three to four steps, three years, and four million dollars. It's a lengthy and expensive process and there are a lot of missed opportunities just because of the length and the cost of this process. What we found out is that each of these steps is a single point optimization, so a single property optimization. You first find molecules that bind to the protein, then you make sure that they can be synthesized, and then you make sure that they can be orally available. We are using quantum computers to do a multi-object optimization so instead of looking at a thousand molecules every time, we're looking at a much larger virtual chemical space, billions of molecules, and find molecules that bind to the protein, easy to synthesize, and orally available at the same step. So instead of doing three different optimizations, we're doing only a single optimization but a multi-object optimization and that's something we can do because we are using a quantum computer, we're using the d-wave annealers, which are very good for optimization problems.

Rene – Got it. Well, that's already answering probably my second question. But the process you described is also known as protein folding partially?

Shahar - You can use protein folding there. We are less interested in those kinds of things of protein folding. We prefer, at least at these stages of the company growth, to work with cases where the protein structure is known. So, you look at the protein energy, it's already folded. Protein folding is an area that in the last year has been through an amazing thing, both on the experimental side and on the technological side with work at Deep Mind and META and some other companies that are doing that. They're using AI, machine learning, and natural language processing to do those parts of things, and in many cases, we use their outputs. We look at the outcome of what they're doing and they support what we're doing.

Rene - Got it. So that that makes a lot of sense. I like you're already alluding to things like a deep fold from Deep Mind and so on, you know, that is using machine learning and AI basically to escalate the process and I think you know like the challenges the industries are facing is computational resources and these are like what you describe, all these, like multi-dimensional optimization problem basically. You have this huge search space and how can Quantum Computing help in that Realm?

Shahar –"So, you know Quantum Computing is a big name. What we're doing is, we are using annealers, and annealers are really good at solving optimization problems. And this is what we're doing, we're using an annealer to solve a multi-objective optimization problem for a specific protein and the specific set of properties that you're interested in a molecule, so that that molecule will become a good drug, less toxic, easy to make, soluble, all of those set of properties. We know how to build a virtual chemical space, large chemical space, billions of molecules, and how to translate that chemical space into a cubo formulaism, and that's what we're running on D-Wave. So, we know how to translate chemical space to cubo, how to translate the constraints that we have, that are in the chemistry language, to cubo constraints, and then we run it on D-Wave. The constraints at the cubo and the constraints are solved using their algorithms. Okay, so they're using their algorithms, we're using our algorithms to translate chemistry into cubes. Right?

Rene - Understood, so basically, you're the translator from the chemistry into the quantum chemistry world. Basically, right?

Shahar - Yes, and translating it back, you know, what's the low energy state coming out of the cubo? You still need to translate it back into chemistry, into how does your molecule look like? What are the properties of your molecule? And how good is it going to be?

Rene - Awesome, and with this, you can also build these personalized medicine, I guess, like you have a person for example, well, unfortunately, I don't, at least I don't, I'm not aware that I have one of a certain disease, maybe I have, I don't just don't know, but you know, potentially, like let's just make this up, like if a doctor finds a certain problem like in my body or something, and there is a certain chemical property or certain protein that, like you said, that is not well behaving, and then they would take a sample and then analyze it and make a very specific drug just for one person. Right?

Shahar - Yes, this is still work in progress, so this is still not something that can happen today, but even today we're looking at smaller and smaller patient populations. One of the projects that we are doing and that we published on is triple negative breast cancer. So, if you look at breast cancer, there are three known mutations, but then there are patients who don't have any of these three mutations. They are the triple negative patients, and they are a very small percentage of breast cancer patients, but we can find and we're working on projects that are targeting that small population now, because we don't need three years, four million dollars to get to these molecules. We're way faster than that. We can look at smaller patient populations, because the cost of drug discovery, drug design is expensive, and if you're doing it faster and cheaper, you can target smaller patient population. We have shown that this specific project it took us, we collaborate with another company, and this took us about six months, the project from the beginning till the end. We're waiting now to see some of those results.

Rene - Wow, and that is impacting life for a person right, like saying like instead of like four or five years, six months, or I mean like you're saying it's just the beginning, right? So it will get even faster and in the long term, and that could actually mean life or death for a person. And so this is amazing.

Shahar - Yes and Quantum Computing, the possibilities there for some problems. Okay, we're not going to make a, you know, a faster Word document processing but for some problems the capabilities of the quantum computer can change especially in science. So when we talk about finding the best molecules for it that will become a drug that's one of the things. When we're thinking about how you design that molecule specifically. Right now we're using annealers, when a gate-based computer will be available that is bigger we can make other parts of the process even faster as well and this is something that we're really looking forward to.

Rene – Yeah. I can totally relate to that. Like, you're saying that's also what I keep on telling folks when they ask me oh when will this be a Quantum like I'm holding up my smartphone here when will this be a quantum computer inside of here and then I tell them well probably never because you should rather think about a Quantum or a QPU, a Quantum Processing Unit a similar to a GPU which is a graphics Processing Unit, an accelerator for very specific problems or what we're also seeing these days and certain computers these VPUs like vector processing units or Tensor Processing Unit super well suited for AI workloads. We're going to see these of course you know quantum computer might probably never be that small we'll see maybe at some point but definitely it's meant for very specific problems but yeah like you're saying these specific problems are huge and impactful for a lot of people like using drug Discovery but also fertilizes all the chemical simulations or optimizations and then even scheduling problems, like you know saving time for a lot of people and then also save and reducing carbon footprint and or running climate models and all of that stuff like where we're reaching really the boundaries of classical computers because you have these multi-dimensional problems and so exciting!

Shahar - Yeah it is! It is isn't it? On the D-Way annealers, there are some really interesting solutions that are being developed now. One of them is the, in the Port of Los Angeles, again cutting down waste, making sure that you know ships don't waste too much time, trucks all of these things, so really interesting solutions.

Rene – Awesome! Yeah, any other solution you would like to mention in the quantum inspired computation space?

Shahar - So there are some interesting work done now with cargo, how you put boxes into airplanes right, that's the Airbus Challenge and there are some interesting works coming out of that as well. And both of these are examples of using the d-wave annealer, you know, on solving day-to-day problems.

Rene – In-fact while you mentioned, it was quantum challenge, like, one of my colleagues who actually took part in the Quantum challenge made second place back then and yeah I actually talked with one of my colleagues in the previous episodes of QuBites about it and yeah it is impressive. If you think about the problem, it's a huge optimization problem right? Like you have a plane that has certain physical constraints it needs to fly in the end.

Shahar - And fly balanced right you can put all the heavy stuff at the front or on the side right, but right they are balancing and this is a multi-object optimization.

Rene - Well this is this is pretty amazing and I I'm really impressed like especially in your field like you're saying it was personalized precision medicine and you know bringing down drug discovery like the face of drug discovery, you're working on like the beginning or the initial phases before the clinical tries like bringing this down to a few moments basically from years. This will be so impactful. Also, if we think about Quantum in it and also AI, I mean when a lot of folks they hear about AI, “oh no you know the robot sare going to kill us”, that kind of a thing but then I always show them like there are so immense progress also for computer vision like detecting melanomes, like skin cancer, which can even reach now a level that is on par with humans or even better to a certain degree and this will scale and will enable this fantastic treatment to many more people in the world. And this is so exciting.

Shahar - Yes, and you don't have to be there in order to employ AI right or machine learning. You don't have to send your Physicians to every part of the world, they can, you know, you can put it on a computer and that computer will reach areas that otherwise may not have access to that level of experience, right? So I think some of that technology is revolutionizing healthcare now not not in 10 years and I think it's really exciting to beat this point now being able to see, to participate, to contribute to those kind of revolutions.

Rene -Yeah, I couldn't agree more. And what was also very insightful in the conversation we were just having that you're already basically using the state latest state of technology with annealing from d-wave with annealer, they have which is almost 6 000cubits or five thousand seventy something or I don't recall exactly but I like you're saying for very specific problems though. So it's not the gate model which is more generic but you're already employing this for, you know, making a super impactful things there and I'm very impressed!

Shahar – Yep, thank you very much!

Rene - Well folks we're already at the end of the show. Thank you so much Shahar for joining us today and sharing your insights again. I'm so impressed and the work you're doing is life-changing for a lot of people. So I'm very much appreciate that you took a little bit of your timeout to talk with me today, thank you so much!

Shahar - Thanks Rene and always happy to talk with you!

Rene – Awesome! And thanks everyone for joining us for yet another episodes of QuBites, your bite-sized pieces of quantum computing. Watch our blog and follow our social media channels to hear all about the next episodes, subscribe to our YouTube channel, you might get a notification and of course you can always visit the website to review all the previous episodes from season one to season six now. Take care, stay healthy and see you soon!

[Check out this article published in ARS Technica, where Shahar Keinan talks about their work for faster drug discovery with Quantum Annealing systems.]