How Generative AI Finds Cures for Rare Diseases Faster

Welcome to episode 063 of Life Sciences 360.


We dive deep into the groundbreaking role of AI in healthcare and life sciences with Shweta from Google Cloud. Learn how AI is accelerating drug discovery, enhancing patient care, and revolutionizing rare disease research. If you're passionate about the intersection of technology and health, this is a must-watch conversation!

🔗 Connect with Shweta on LinkedIn: https://www.linkedin.com/in/shwetasingh01/ 
📍 Explore more about Google Cloud: (https://cloud.google.com)

Chapters:

00:00 Introduction  
00:01 Using Generative AI for Drug Discovery and Molecular Structures  
00:11 Synthetic Images from Histological Data for Rare Diseases  
00:20 Bayer's AI for Regulatory Dossiers  
00:38 Categories of AI Perspectives in Healthcare  
01:00 Shweta’s AI Journey & Joining Google Cloud  
02:01 Methodical Approach to AI in Healthcare  
02:43 AI for Rare Diseases & Accelerating Drug Research  
05:01 First Steps for Companies Embracing AI  
06:51 Real-World AI Use Cases: Ginkgo Bioworks & Bayer  
10:32 Potential Future Frameworks for AI in Regulatory Submissions  
13:52 Challenges of Rare Diseases & AI Solutions  
16:05 AI's Role in Accelerating Research & Repurposing Drugs  
18:10 Broader Applications of AI in Life Sciences Beyond Rare Diseases  
19:22 AI for Drafting Clinical Trials & Patents  
21:27 Addressing Concerns of AI in Healthcare: Ethics, Safety, and Bias  
24:46 Human-in-the-Loop: The Key to Safe AI Use in Medicine  
30:39 The Future of AI in Healthcare: Democratizing Access & Proactive Care  
37:06 Outro & How to Connect with Shweta

---

Links:

*Harsh Thakkar LinkedIn ( https://www.linkedin.com/in/harshvthakkar/ )
*Listen to this episode on the go! 
🍎Apple podcast: https://apple.co/3RXPoS1
🟩 Spotify podcast: https://spoti.fi/3EbDZbr

👍 Like this video if you enjoy diving deep into the latest healthcare trends.
💬 Comment below any questions you have. Have you implemented this into your operations yet? 
🔔 Subscribe to our channel @lifesciences360 for more expert interviews about the rapidly evolving world of life sciences.
📲 Follow us on social media to keep up with the latest news and discussions in the biotech and pharmaceutical sectors.

Thank you for watching and supporting our channel!

#AIinHealthcare  #LifeSciences #DrugDiscovery  #GenerativeAI  #ArtificialIntelligence #HealthcareInnovation  #RareDiseases  #AIMedicalResearch  #PharmaTech  #HealthcareAI  
#PatientCare  #GoogleCloud  #AIDrugDiscovery  #TechInHealthcare  #HealthTech  
#FutureOfHealthcare  #MedicalAI  #Biotech  #DigitalHealth  

For transcripts, check out the podcast website - www.lifesciencespod.com

Transcript

Harsh Thakkar (00:02.095)
Shweta, you have been at the forefront of the whole AI revolution that's going on in healthcare and life sciences. To set the stage, why do you do what you do? Or what do love about it?

Shweta Maniar (00:18.52)
Well, you know, I think part of the conversation that we're having today, right, is really going to be set around the concept of rare diseases. And I think that's a big part of why I'd like to spend some time articulating this relative to rare diseases. Rare diseases by their very nature present a unique set of challenges when it comes to the research for rare disease, the diagnosis for rare disease, and then the cures for rare disease.

And that's because there's limited patient populations, scattered medical knowledge. There's a lot of diagnostic complexities that make that a really tough nut to crack. But I believe that AI and specifically generative AI actually holds an immense potential to transform some of that landscape. And this is actually why I do what I do and spend the time at these crossroads of the

industry, healthcare and life sciences, and how do we apply the technology to serve this industry? So, go ahead.

Harsh Thakkar (01:21.591)
Yeah. No, I was going to say when I talk to people about AI, there's basically three categories. One, I'm bullish. One, I don't care. And one, I kind of know it's useful, but I'll wait until somebody tells me to do something about it. I know which category you are in, but I want to know why. Or when did you make that switch to be in that category as a bullish?

Shweta Maniar (01:50.818)
Yeah, I think that, so prior to joining Google, I had the opportunity to be in the healthcare industry, right? And so this was working in hospital systems, working in the pharmaceutical organizations. And I was on the receiving end of these different technologies. And I think part of when joining Google, right? Part of this is understanding here's where we can go, right?

and also then understanding what we are doing today and where can technology actually derive value today. So of course I'm going to be bullish, but I'm not going to say that this is going, that tomorrow we are going to, today we are going to be able to create a medicine from scratch just overnight, or we are going to diagnose no humans in the loop. It's very important to look at this at a much,

methodical way. And when you're looking at this in a methodical way, you can identify lower hanging fruit, your medium term goals and your longer term goals and how you can apply AI. So I'm bullish because I was on the receiving end of pitches, here's your silver bullet, you know, it's going to solve everything. And I will not I will not share names of organizations. Whereas in actuality, harsh, it really is. Here's the technology.

but we need to work together to build and apply these technologies, right? When it comes to rare diseases, again, coming back to rare diseases, you can accelerate drug discovery, right? Or you can accelerate research, which we all know is like a costly endeavor, but, and you can use potentially generative AI to accelerate the processes because you can use that to accelerate.

Harsh Thakkar (03:19.577)
Mm.

Shweta Maniar (03:43.95)
predicting drug interactions, right? You can optimize molecular structures on your screen instead of at the bench and wait for days or months. And so there's a lot of ways that generative AI has the ability to support the rare disease space, but we have to be methodical about it, right? This is not going to be something where we can look at the KPIs tomorrow. There are other things that AI can do for today, right? So I remain bullish because there are things that you can do today, empower patient communities. You can use AI.

Harsh Thakkar (04:05.198)
Hmm.

Shweta Maniar (04:13.858)
to otherwise isolated patient groups to find relevant information or to find the right support creating virtual communities or chat bots or offering somewhat personalized guidance. That's something that can happen now and in a much efficient context than to say I can make my medicine tomorrow. And so there are different gradations of the way you apply AI in a space.

Harsh Thakkar (04:41.453)
Yeah, I love two or three huge takeaways there, which is one, there has to be an approach, a methodical approach. And whether you take that from some external sources or standards like NIST or something else, or whether you define what that means for your organization. And then the second that you said, is having that vision, you're not gonna overnight come up with something that's gonna change everything for you.

But I wanted to dig deeper on this because you've told us that the potential is huge, especially for rare diseases to help with accelerating the research or improving the diagnosis. But for companies who are looking to take that step and start using AI, you also mentioned that it's easier to do it on your screen versus spend time in the lab trying to do it the hard manual way. For companies that are looking to take that step,

Do you have a three -step approach or or maybe what is the first step that you want them to take before they do anything else if they are in going in this direction?

Shweta Maniar (05:50.146)
Yeah, I think one of the first steps, to take for anybody, doesn't matter what industry or what part of the healthcare life science industry that you are a part of, it is really around first educating yourself. I think there's a quote from Stephen Hawking that intelligence is the ability to adapt to change. And so when you look at generative AI, it isn't just another tool, it's actually something that can really change and transform the way you diagnose diseases and develop

Harsh Thakkar (06:03.46)
Hmm.

Shweta Maniar (06:19.886)
treatments. But in order to do that, right, this is not just those who are going to be your technical experts that need to understand part of how we're going to adapt and build trust in AI is going to be around educating yourself and making sure that you understand its implications and keeping the human in the loop entirely, but also educating everyone across the board.

This is not just about making sure that it's just your technical experts who understand how the AI works. But if it is going to impact how we all work in the future, it's actually incumbent to step one is that we need to build an understanding across, even in a layman's terms. And then two, being able to trust into that system, trust into that AI that if you're putting a hundred things down in the funnel and it's recommending for you to look at these 10 molecules, why did that happen?

And so I think part of it is education building trust into the system before you even can get started.

Harsh Thakkar (07:13.07)
Hmm.

Harsh Thakkar (07:19.413)
And, you know, as you are at Google, which is, one of the most forward thinking technology organizations in the world with many others in that category. So I'm guessing when these types of conversations are going on, you probably have more real world examples or some experiments that you've tested within the company. Can you share any examples that you've seen where generative AI has

potential or has made a big difference either in research or patient care.

Shweta Maniar (07:52.906)
Yeah, I think there's two really good examples that I think are really fantastic to share for today. We, from a Google perspective, have a partnership with Ginkgo Bioworks. Ginkgo Bioworks, they're developing pretty powerful large -scale language models, which are the fundamental building blocks of life, proteins and DNA. And these models actually open the door to a much wider range of possibilities.

You could be designing your therapeutic proteins and even industrial enzymes to create some of the regulatory sequences that you need for gene therapy and other applications. Ginkgo's vision is to empower the biotech ecosystem. Rather than developing the drugs themselves, they're actually looking at how do they make their models accessible to partners through APIs, through software packages, through different types of collaborative partnerships.

And so by leveraging Google cloud, Ingo has been training, refining, and deploying a lot of these sophisticated language models for a diverse range of tasks. And their goal is to use this generative AI to create comprehensive representations of proteins and DNA across a variety of different organisms and cell states, which hopefully, right, is to lead to the development of a

a new versatile set of tools and have broader applications. So that's Ginkgo, right? When we're talking in a much more biotech context. Another example that might be a little bit more tangible for some of your, some of the other audience members is our work that we do with Bayer. Bayer is using AI to improve some of their drug discovery processes by analyzing different sets of data and automating a lot of tasks for insights and speed and, you know, to speed up some of their new medicines.

So we work with Bayer across a wide variety of technologies with the ultimate goal to help them support from kind of from the developers, from idea to deployment and bring forward new AI enabled solutions or applications. Generative AI specifically has actually enabled Bayer to automate the population of 70 to 80 % of their regulatory dossiers.

Shweta Maniar (10:20.824)
So that actually is streamlining the regulatory process, which ultimately, right, the goal is to then, if you're able to move the paperwork faster and more efficiently with the same level of accuracy, you would like to speed up the availability of medicines that ultimately are going to the patients. Also with Bayer, I'll give you one other example is the use of synthetic data.

Harsh Thakkar (10:20.942)
Hmm.

Harsh Thakkar (10:47.907)
Hmm.

Shweta Maniar (10:48.286)
In pharma, they're already working on synthetic images for oncology that are actually created from histological images. And that's really important in the rare disease space because of the limited data that exists on training algorithms. So this is a really important use case, particularly as we're talking about rare diseases.

Harsh Thakkar (11:10.177)
Yeah, and as you were talking about this, one of the examples that came to my mind also on this topic was, again, this is just one of the projects that we've been working with, Big Pharma Company, and we were having this internal meeting, and they mentioned that there is a potential maybe in the next three to five years somewhere where all the regulatory agencies are gonna maybe

come together and decide a framework of how the regulatory submissions should be made. Because currently there is a mix of technology, but there's a lot of human reviewers at FDA and bunch of other regulatory agencies in the world. But how can they decide what component of the submission should be in a machine readable format versus human readable? And then if they decide, hey, this is what the new rule is,

how do companies, they're gonna be playing catch up to get that data in a machine ready state. Whereas right now it's still, nobody has made that decision. There are some companies who are thinking along those lines. Have you heard anything? I'm just curious.

Shweta Maniar (12:23.906)
Yes, absolutely. think that there are individual organizations that are looking at, know, it's not just around the application of AI, but there's also this angst, right, about how do we apply AI and how do we know the regulatory bodies will accept it? And then there's many consortiums that different pharmaceutical, biotech, medtech, medical device companies are creating in order to create some sort of standardization. And simultaneously, we have

Harsh Thakkar (12:39.886)
Hmm.

Shweta Maniar (12:53.07)
We have regulatory bodies like the FDA who are then looking at how do they inform themselves on understanding the implications and the applications of these types of technologies because they know that they need to inform themselves in order to create a point of view and guidance to the industry.

The short answer is, know, we as Google have the opportunity to support the education to some of these regulatory bodies. But ultimately, what I've been very pleased to see is regulatory bodies are not necessarily creating a, you know, a wall of, hey, you know, no AI. This all has to be done the old way. I think it's really, we've come a long way where everyone is trying to educate themselves and understand this because with the evolution of AI moving so quickly,

I think that it's really been incumbent upon both the industry as well as the regulatory bodies that have to provide the guidance. Both are trying to understand and create their points of view so that they can adopt these new ways of working at the same time, create the same level of safety and trust in these systems.

Harsh Thakkar (14:08.047)
Yeah, and absolutely for us as a service provider and a consulting agency, we are like the facilitator between the companies and the regulators on many of these projects. So it's extremely fascinating time to be working in life sciences and seeing all this going on and just sitting and thinking how it's all going to play out in the next five or 10 years. It's going to be amazing.

Shweta Maniar (14:33.272)
Yeah, absolutely.

Harsh Thakkar (14:35.511)
I wanna go switch gears into and go into rare diseases. At the top of the episode, you mentioned that rare diseases are difficult, it's tough to crack exactly what's going on. We've talked about AI and the potential there, but if we are to switch into the direction of rare diseases, what are some challenges that make it hard to crack rare diseases?

Shweta Maniar (15:04.952)
Yeah, you know, I started out, right? Part of it is around data, amount of data, data access, right? By the very nature of rare diseases, means that you have a disease affects a small number of individuals. So that makes research and development incredibly difficult.

If you think about it, right, with limited patient populations, gathering enough even data, let's just talk about it in the context of clinical trial, it's a major hurdle. Plus with the rarity, right, it's often you've got some vague symptoms, maybe lengthy diagnostic, know, delays, you know, there's the emotional part that certainly impacts patients and their caregivers and their families. And then to top it off, you've got all these complexities around

the conditions, what they mean, what kind of treatments are available, making your decisions. So AI actually offers some hope here, I would say, because you can use technologies like generative AI to create some synthetic patient data to augment some of the limited real world information that we have. So it's like, it's just like expanding the research data with, know, expanding the research pool without actually getting more patients.

Harsh Thakkar (16:27.342)
Hmm.

Shweta Maniar (16:27.576)
So AI algorithms can be trained for faster diagnosis, for earlier interventions. I've even seen some organizations who are looking how can they repurpose some of their existing drugs. AI can actually sift through existing medications and identify some patterns that wouldn't otherwise be caught or picked up on for potential new uses in the context of rare diseases. So it's like,

Harsh Thakkar (16:41.891)
Hmm.

Shweta Maniar (16:54.572)
It's like finding a hidden treasure in a library of existing drugs. So when it comes to, mean, you can see I'm getting really excited when it comes to the rare disease space, because it is a huge, huge challenge. And the innovative application of a lot of these AI tech, of course, especially, you know, platforms like Google Cloud, it really gives us a good, good opportunity to help.

and it helped a particular part of the industry that could benefit from the application of these technologies. And I think it's a testament to the power of how AI can really help the healthcare and life science industry and also provide additional hope to people who are living with rare conditions.

Harsh Thakkar (17:47.275)
Yeah. And you mentioned one important point, which, you where you said that AI is going to make it better by providing, you know, synthetic, like learning about the data and then creating other data sets and, you know, maybe giving the researchers an angle or a question to pursue. We're not saying that AI is going to come out with the magic answer, but

AI is going to be able to process data much faster, maybe present new angles or new areas to look at, or maybe study different treatment plans, maybe combinations of outside of... So those are all for rare diseases, but do you know of any other areas outside of rare diseases where you think AI could also help?

Shweta Maniar (18:39.756)
Yes, you know, I think that if we're to pull away from the use of AI, right, or the use of AI, particularly in rare diseases, our conversation has been of course focused on rare diseases because of the scarcity of data, but really, you can apply this to the entire life science industry, right? It is, you can use this in areas that are for research and being able to make sense of different data types to be able to

have a more holistic understanding of an individual because harsh, know, if you have the same disease as I do, our genetic makeups, our environmental conditions, our lifestyle, everything is going to have a different, you we're going to have different outcomes. But also there's also the use of AI, generative AI in a more, what I would consider a lower hanging fruit opportunity when it comes to supporting the healthcare and life science industry to perhaps

some automation of high administrative burden opportunities, whether that is in research to collate a variety of research types of research data, or what is the latest research on this and being able to sift through and contextually give you information on scientific data. It could also be in a clinical trial context of using AI to help graft your

Harsh Thakkar (19:44.484)
Hmm.

Shweta Maniar (20:07.15)
first version of your clinical trial protocols. I have seen the application of generative AI in life sciences to draft the first version of patents. Again, these are all with the human in the loop, right? This is not going to give you your patent from zero to 100. But the idea is now you could use your skilled employees not to draft that first version from scratch, but actually giving

Harsh Thakkar (20:17.731)
Hmm, interesting.

Shweta Maniar (20:32.438)
documents again, it could be any document, any part of your value chain in the life science industry that is heavy on paperwork and documents is an opportunity where you can apply generative AI to create your first version that then you can correct, improve upon as opposed to starting from scratch. And we've seen the application of these types of this technology in these type of applications being used by life sciences more and more.

and healthcare, but more and more because of just the shorter time to derive value, right? If you, know, once you start leveraging this type of technology, now you're able to show time savings. You can show the start and end of, you know, finalizing a protocol from zero, from scratch or from zero to when it's fully ready to go. We're able to now.

work with organizations who are showing some statistically significant reduction in time or being able to reallocate their employees, et cetera. So there's a lot of applications outside of the rare disease space where you can see the usefulness of AI in life sciences.

Harsh Thakkar (21:33.134)
Hmm.

Harsh Thakkar (21:45.335)
Yeah, I love the one about patents and having that first draft and then bunch of other examples that you share. I've also seen many others. And again, the common theme is AI is not going to substitute the human effort, but it's only going to augment it. And the human still has to do the critical thinking in this whole workflow and then use AI as like a technology to augment what they're doing.

Shweta Maniar (22:02.222)
Absolutely.

Shweta Maniar (22:15.278)
Absolutely.

Harsh Thakkar (22:16.589)
So I know, as I mentioned at the start of the episode, are, you know, there's sort of like three camps of people in this whole AI side where obviously, you you mentioned you're passionate about it. You've shared so much. And there's this area of people that are very hesitant to get in and some that are in the middle that are like, okay, I want to get in, but what about X? What about Y?

So to them, the X and Y is, what about patient safety? What about the ethical and regulatory implications? What about bias? What do you have to say to those people who are not...

Sorry, I just got a pop up that your recording was stopped. Let me see.

Harsh Thakkar (23:16.215)
Okay, I think it's going on. I don't think there's any issue.

Okay, so if it's okay, I can start just by the question, what do you have to say to, yeah. So what do you have to say to those people and how are you in your role making sure that you're balancing innovation with all these other concerns?

Harsh Thakkar (25:56.334)
Yep.

Harsh Thakkar (27:08.174)
Yeah.

Harsh Thakkar (27:34.871)
Yeah, you made a really important point about monitoring and continuous improvement. Maybe that's not something in the traditional approaches, which are more static, where you've complied with some requirements and you know it's working. If it ain't broke, don't fix it, that kind of mindset. But that really doesn't work when you start bringing AI because it can process information so fast.

you constantly have to make sure you're not deviating from your end goal. So that is a huge challenge and I don't have the answer of how companies are going to monitor that and make sure it's heading in the right direction, but that's a really good point there.

Harsh Thakkar (28:19.807)
So you talked about collaboration and making sure that we're keeping the patient at the center. So you've had an amazing trajectory in healthcare and life sciences. You're currently at Google, you're doing amazing stuff. So what advice do you have for other leaders in healthcare and life sciences to maybe either start these conversations internally or maybe partner with other organizations just to see how we can collaborate together?

Harsh Thakkar (30:10.306)
Mm.

Harsh Thakkar (30:52.26)
Mm -hmm.

Harsh Thakkar (31:27.919)
Yeah, definitely. I'm with you on start small and then learn and scale, but don't wait to start, is, I think every day with a new development in AI, think it's signaling to us that don't wait to start. Start because the use cases are plenty for organizations to, So looking at the future of AI, maybe five or 10 years ahead,

What is it that's most exciting to you? you do you are you hoping to see something? Are you hoping to be surprised by something? That's not yet out there, but Anything you want to share there?

Harsh Thakkar (32:23.31)
Hmm.

Harsh Thakkar (32:38.585)
Hmm.

Harsh Thakkar (34:39.597)
Yep. I think, you know, for anybody that's listening or watching this episode, you know, there's a bunch of takeaways here that everything we've heard Shweta say today. The first one that really stuck with me is that, you know, there's always going to be a human component in whatever you're doing with AI. So don't have the fear that, you know, AI is here to replace humans, but try to educate.

Try to learn and if you're a leader, try to make sure that you're giving the training opportunities. If you're an employee, maybe ask your employer to learn about AI and don't just wait because it is going to impact your role in some way. And just try to learn how you can work with AI alongside AI. So that's one takeaway. The second one was because it's a new technology, it's going to need certain direction. It's going to need certain...

amount of observation and monitoring. And that's a great point that we touched upon. So if you are using AI, make sure how are you gonna keep it on track? How are you gonna let AI take you to your north star if it's a patient or whatever you're trying to achieve? And then the third one that you mentioned is don't wait to start. Look at opportunities, look at things in your organization, low risk projects where you don't have to spend a lot of money.

try to use it, if it breaks, try to learn from it, and then slowly expand. So think those are the three big ones for me here today.

Harsh Thakkar (36:29.337)
Hmm.

Harsh Thakkar (37:25.644)
Right.

Harsh Thakkar (38:18.479)
Yeah, 100%. And listen, thank you so much. It's been amazing having you on here and learning about everything that you're doing in your role and all the exciting projects that you're working on and sharing with us, you know, why you're passionate about AI, what do you see it doing in the life sciences field. For any of our listeners or viewers, if they want to connect with you after this episode or maybe, you know, learn more about you, do you want to share your social media?

where they can get to you.

Harsh Thakkar (39:05.451)
All right. That's it for today. Thank you, Shweta. Appreciate it.