Scroll Top

Gini Deshpande of Numedi on Augmented Intelligence for Drug Discovery


This week Harry talks with Gini Deshpande, the co-founder and CEO of San Mateo, CA-based NuMedii, a company making judicious use of big data and AI to speed up drug discovery.


Gini Desphande says she likes to think of “AI” as augmented intelligence rather than artificial intelligence: a system of human plus machine intelligence that can speed up drug development and cut R&D costs and failure rates in clinical trials. AI “really isn’t at the point where it’s automatable,” she says. “We still need a lot of human intelligence to be coupled with this technology, to determine what are the questions you want to ask and to evaluate all the targets that come out, to say ‘Do these make sense?'”

NuMedii’s specialty is analyzing bulk tissue to isolate gene sequences in single cells that can point to new drug targets and drug candidates for diseases such as idiopathic pulmonary fibrosis. “The AI component helps us look at new targets that are not obvious to the human eye,” she says. “It enables us to find network-level connections between diseases of interest and targets that are relevant for that disease. We can look at which nodes are coming into play and which ones should be manipulated for a particular disease.”

This episode is part of a special series of interviews with speakers at the AI Applications Summit produced by Corey Lane Partners. Check out the full show notes and other MoneyBall Medicine episodes at our website.

Please rate and review The Harry Glorikian Show on Apple PodcastsHere’s how to do that from an iPhone, iPad, or iPod touch:

1. Open the Podcasts app on your iPhone, iPad, or Mac.

2. Navigate to The Harry Glorikian Show podcast. You can find it by searching for it or selecting it from your library. Just note that you’ll have to go to the series page which shows all the episodes, not just the page for a single episode.

3. Scroll down to find the subhead titled “Ratings & Reviews.”

4. Under one of the highlighted reviews, select “Write a Review.”

5. Next, select a star rating at the top — you have the option of choosing between one and five stars.

6. Using the text box at the top, write a title for your review. Then, in the lower text box, write your review. Your review can be up to 300 words long.

7. Once you’ve finished, select “Send” or “Save” in the top-right corner.

8. If you’ve never left a podcast review before, enter a nickname. Your nickname will be displayed next to any reviews you leave from here on out.

9. After selecting a nickname, tap OK. Your review may not be immediately visible.

That’s it! Thanks so much.


Harry Glorikian: Welcome to this special series of Moneyball medicine focused on AI machine learning and analytics applied to drug discovery and development. This special series was recorded as part of the AI applications summit produced by Cory lane partners. I’m your host, Harry Glorikian. In this series, I will interview different speakers from the event and we will hear their experiences.

We will dive into the challenges and opportunities they’re facing and their predictions for the years to come welcome to Moneyball medicine.

My next guest. Likes to refer to AI as augmented intelligence, more than artificial intelligence. The goal of combining the system of human plus machine intelligence to help speed up drug discovery, cut R and D costs and decrease failure rates in clinical trials, all of which can eventually lead to better, more precise medicines.

Gini  has spent several decades turning, cutting edge life sciences technologies into products for patient benefit. She is the co-founder and CEO of NuMedii, a data-driven drug discovery company focused on using big data and artificial intelligence technology to rapidly innovate the drug discovery process.

As CEO, she has structured partnerships with large pharma companies and raise the company’s initial rounds of funding prior to a NuMedii. She has helped companies identify optimal markets and whole product solutions for their groundbreaking technologies. She has also led innovation within the world’s leading pediatric hospital, focusing on the creation of new devices for the tiniest of patients.

She has also helped commercialize early stage technologies in research tools, diagnostics, and therapeutics, and has closed licensing deals worth several million. Gini has coauthored numerous papers given numerous talks and is the recipient of several awards. She has received her PhD in biological sciences from Purdue university and did postdoctoral work at Harvard medical school.

She graduated at the top of her MSC class from. University of Poona India. Gini, welcome to the show.

Gini Deshpande: Well, thank you for having me.

Harry Glorikian: Great to have you on the show. Let’s start off. Um, Tell the listeners, tell us about a little bit about NuMedii and what you’re trying to accomplish and where, where is it in its life cycle?

Gini Deshpande: Absolutely. Um, so NuMedii has been one of the pioneers in the AI for drug discovery space. Um, and what we’re really focused on is now leveraging our platform, which consists of a lot of proprietary single cell sequencing data. Um, along with other capabilities and other data sources, um, to identify new therapeutics for diseases like fibrosis, uh, specifically we’re focused on a disease called idiopathic pulmonary fibrosis.

It’s a really bad condition that impacts the lungs. It makes it really hard for patients to breathe. And it has a mortality rate that’s higher than some of the cancers combined. So really, really bad disease, um, for which there’s a need for new therapeutics. And so NuMedii he’s focused on leveraging his platform technology to develop new therapeutics in this space.

Um, we have previously worked with a number of pharmaceutical companies and help them with their discovery efforts. Um, so felt that it was the right time for us to start taking these molecules further ourselves.

Harry Glorikian: So when you say platform. Right. I could think of a, you know, lab based platform. I could think of an IT-based platform I could think of.

So when you say platform and we’re thinking data analytics, data, um, analysis for the most part is how do you see you when you’re pitching it a partnership or whatever. How do you position that to your partner and what is the secret sauce without, of course, You know, divulging confidential information.

What is it about n NuMedii that makes it special?

Gini Deshpande: So when I refer to a platform, I’m really talking about our data sciences platform. So obviously there’s various types of data that we collect, uh, some from the public domain, some that’s proprietary and generated either by NuMedii or its partners and collaborators.

Um, so data sort of one component of the platform. And then there’s the algorithms, the AI algorithms that we utilize, um, to, for targeted discovery, for identifying compounds that are relevant for a given disease of interest. So it’s really these two components that, that are that from the basis of our data sciences platform.

So that’s essentially what we leverage. And, um, the secret sauce here really is that by leveraging AI and data such as single cell sequencing data, which now enables us to get very granular and understand which cell types are contributing to a particular disease process. And therefore what targets are in those cells, that one should then.

Uh, modulate for therapeutic benefit, we can really get very precise with the cell types and the targets that we want to impact with our platform. And that’s essentially what the, what the technology enables us to do. Um, the AI component of it really helps us look at targets that are not obvious. Uh, to the human eye.

This is not something that we would, uh, look at and say, oh yeah, I know this target. So that’s the sort of secret sauce, if you will, of what the technology enables us to do. It enables us to find network level connections between, um, diseases of interest. And, and targets that are relevant for that particular disease.

So think of, uh, a network, if you will, that’s coming into play in a particular disease. Um, we can start to look at all the nodes that are coming into play in that, in that network. And then start to look at which of those nodes should be manipulated or modulating for us to actually benefit that particular disease.

Harry Glorikian: Well, so you can actually like start to look at different cell types at different states and be able to create profiles of them.

Gini Deshpande: Absolutely. You can start to look at individual cells. So for instance, in the case of IPF, one could look at macrophages that are coming into play. One can look it up  cells.

You can start to really distinguish between these cell types. Previously, when we’ve looked at data that’s been collected and that’s been in the public domain, it was usually what was considered bulk tissue data. That means somebody took a biopsy. And in that biopsy, you had a mixture of different cell types.

Harry Glorikian: Right, right

Gini Deshpande: And then you profile that using a high throughput biology technology like microarrays. And so you looked at all the changes that were happening at the transcriptomic level, uh, across these, but you didn’t know whether it was macrophages that were contributing to the signal predominantly. Or there epithelial cells

Harry Glorikian: Right

Gini Deshpande: Now, our co-founder of Atul Butte has come up with ways to deconvolute signal from bulk tissue, but it’s still challenging to get very granular and really identify precisely what’s contributing to signal and what targets to look for in those cell types. And so single cell sequencing

Harry Glorikian: Makes it a lot easier.

Gini Deshpande: Really enables us to absolutely, get into that level of granularity now, and then confirm your findings.

So you find it interesting target. You can go back and test it and confirm that it is that target in those fibroblasts, that is that’s coming into play and an important for a disease process.

Harry Glorikian: Sure. Let’s just take one step back here. I know you and I have talked about this whole concept of AI and et cetera.

Right. Um, the nomenclature drives me crazy. So, uh, but where do you think we are and how would you sort of deconvolute the space to a certain degree.

Gini Deshpande: That’s a great question. And, you know, I get asked this often, um, so two things that I will, I will highlight one is that AI, uh, people think of AI as artificial intelligence.

And the reality is the state of the art today is it’s it’s. I would consider it to be augmented intelligence. It really isn’t at the point where it’s automateable and one can make sense of what’s coming out of, of a technology and just be able to run with that as it is. It’s nowhere close to that. We still need a lot of what I consider a human intelligence to be coupled with this technology to really make sense of it.

So you need human intelligence in the front end to determine what are the questions you want to ask? And therefore what data should get fed into the system. And then we also need human intelligence to evaluate all the targets that come out of a discovery platform to say, do these make sense? Have I seen these targets before, are they biologically relevant?

Therefore, when I find something novel that I haven’t seen before, I can start believing in that because you have a level of confidence that,

Harry Glorikian: Right.

Gini Deshpande: The other is that this field, when, you know, since it’s, since Numedi, you got good founded and launched has now evolved in a very interesting way . So, if you think about drug discovery and development, obviously it’s a multi-step process and the drug development aspect of it, which is testing the drugs in patients, um, has an entire sort of, uh, spectrum of companies that are working in speeding up the clinical trials process, uh, improving patient stratification.

All of those pieces that are what I would consider downstream of drug discovery. If one were to look at drug discovery itself, you were starting to see sort of an emergence of companies in what I would consider three major categories. One is companies that are focused on target discovery. Numedii is very much in that category.

And then there are companies that are focused on leveraging AI for speeding up the search of chemical matter. That is relevant against a given target. So let’s say we come up with a new target. We need to go through high throughput screening to find. Uh, the right compounds, right? And AI companies are now focused.

Certainly AI companies that are focused on speeding up that process of finding the right chemical that will modulate that target of interest. So going from, you know, um, target to hit, to lead that hit delete piece is where these companies are coming in and speeding up the process, uh, and, and enabling us to come up with, with compounds in a much faster process than, than we’ve previously been able to do.

And then the third category is what I would consider. Companies taking orthogonal approaches to drug discovery. So phenotypic screening high-throughput phenotypic screening for example, is, is one sort of category where, um, companies such as the recursion pharma are taking advantage of, um, taking cell cell types of interest and applying compounds, and then applying AI to the images from those, from those screens, to be able to speed up the discovery of, of relevant uh, molecules. So you th you have three distinct categories of companies that are all focused in the AI for drug discovery realm, if you will. Um, and they’re, they’re not necessarily competitive with each other. They’re quite complimentary in some regards, if you look at, you know, what Numedii does and what a company that’s focused on, the chemistry aspect of it does, we will be complimentary to each other.

So I think we’re starting to see this specialization, if you will, within the sector where companies are playing to their strengths.

Harry Glorikian: You know, I say sometimes I think if a big pharma has sort of thought this through, they might actually put these links of the chain together and completely be able to rethink the drug discovery slash development process and make it hopefully shorter and theoretically less expensive.

I’m waiting to see the fruits of all that labor, but at least what I’m seeing right now looks like. We’re headed in that direction.

Gini Deshpande: I would like to, I think so. I think, you know, it’s, it’s one where, um, the sense I get as pharmas in this wait and watch period, they are embracing AI technologies. It’s a little bit, uh, of a slower adoption curve, if you will, because, uh, they have some skepticism.

Rightfully so. Um, I think once we have a few success stories coming out of this field where you see the impact in the clinic, where I think that’s where we really need to have, uh, the ability to demonstrate that these sorts of technologies are going to have a meaningful impact. Uh, we’ll start to see broader adoption of, of these sorts of, um, technologies and capabilities.

And certainly pharma companies are doing a good job of embracing AI and incorporating it into their, into their workflows.

Harry Glorikian: That’s where I’m I get skeptical, right? Because I talked to a couple of people that get it right, but then there’s this 10,000 or 20,000 person organization. And there’s not 20,000 people that understand what’s going on.

Right. It’s five, eight. ten. And so I always feel like there is a Kodak moment coming where, you know, if, if you were able to find, you know, prove your target discovery engine, do you really need the rest of, you know, the value goes through the roof in a sense if you think about it. Right. So I think there’s a timing of when they move into it.

The other part of it is, is I’ve noticed. It’s not just pharma, that’s interested in this, these engines, but tech is also starting to show an interest saying I’m interested in the data analytics engine, part of it, not the wet chemistry and stuff like that, but I’d like to provide it as a service. So I’m, I’m looking at this and saying, I feel like the, the, the landscape is changing.

Um, you and I talked about Amazon at one time. And how they’re thinking about you think they’ll eventually be able to, you know, want to develop drugs?

Gini Deshpande: Absolutely. I think, you know, um, two things here, right? One is that the value proposition in the biopharma industry still is a clinically active molecule.

So once you’ve proven that your drug works and it works better than anything else, that’s out there. Um, that’s where you’re going to have the aha moment. We are not there yet. There’s a number of drugs that are in clinical trials right now. Hopefully we’ll see, you know, many more coming into the, into the clinical pipeline, but we’re not at that point yet where we have enough number of molecules that have gone, that have been discovered using AI technologies that have gone through clinical testing and therefore shown.

That you are improving the probability of success. And by the way, just as an aside, I’m not convinced that AI, as, at this point in time is necessarily going to significantly save time or cost.

Harry Glorikian: Right

Gini Deshpande: I think it really boils down to a value proposition of improving the probability that your drug is going to be more effective and that’s sort of where I think there’s going to be tremendous value.

Um, so I think that the field is still early enough that we don’t have a significantly high enough number to be able to come up with any sort of stats to say, you know, AI technologies are better than the tried and tested. Uh, historical approaches that have been utilized to come up with with new new targets and therefore new molecules.

Um, to your second point about tech companies getting into the space? I, again, get back to my earlier point. I think that value is in the drug candidates themselves. So while there is a need for companies to provide services and there’s always historically been a need for companies to provide services to the pharma, to the biopharma industry.

I think the value proposition still remains in the drug candidates themselves. And if you look at the drug discovery development process, and particularly in companies, such as ours, that utilize the data sciences approach, we do take advantage of the commoditization if you will, of cloud computing. So

Harry Glorikian: Right.

Gini Deshpande: Amazon’s already at one end of the spectrum, um, through their acquisition of PillPack there, they’re going to be at the other end of the drug discovery development process, which is providing the medication to the patients.

Harry Glorikian: Right

Gini Deshpande: My, my guess, your, is that something where along the way, companies like Amazon are going to want to get into the middle of the, of the value chain, because they can probably add something to the process and probably help contribute in a significant way.

Um, so I think, you know, 10 15 years from now, can Amazon be a major player in this space? Depends on what their, their core focus and interest is. But you could imagine companies that are providing, you know, cloud computing and actually are in close contact with the patients through delivery technologies, like PillPack could actually come in and, and contribute even more to the story.

Harry Glorikian: Yeah. If you think about the, in the last five years, how much computing power is become available at the time? Flip of a switch or a push of a button. I mean, it’s, it’s unbelievable. And I don’t see it slowing down. At all.

Gini Deshpande: Absolutely. And I think it’s just, you know, it’s going to get better and better with time.

So I certainly think that the costs are going to come down. So our compute costs that are going to be significantly lower, which is going to enable us to do a lot more than we’ve previously been able to do. Um, with data. And so just the sheer volume of data. And what we want to do with it is now, uh, particularly in biological sciences where it’s still not at the scale of like, you know, data from Facebook, for instance, it’s still smaller in terms of scale.

If you look at just the sheer volume of data, but the ability to utilize. Um, AI analytics now on top of that kind of data, certainly very feasible through, um, you know, the availability of cheap or cost-effective cloud computing.

Harry Glorikian: Well, and now, and now more and more, more and more groups are actually generating data that’s useful and the combination of different augmented intelligence techniques coming together.

I’m seeing sort of that next. Curve being created of opportunity. I would say the next I’m looking at the next five years and going, you know, I feel like every few months I’m going to read some new, very interesting story, which is all by itself. Probably not fantastic, but in a string of stories, And an event that it will create this very interesting dynamic that’s happening in the market.

Gini Deshpande: I actually would take it even further and say, you know, if one were to let our imaginations run wild, what would the next Genentech look like? I think it will be a data sciences company. Yeah. Because of the availability of data, the ability to generate more data proprietary data. The ability to go deep into certain areas, the ability to profile biology at unprecedented scales, um, the reduction in cost.

And so to some extent, by being able to take advantage of all this data and come up with a number of targets that can then be quickly validated. And as you mentioned earlier in our discussion, if you can string those, those pieces together and have a complete. Um, you know, end to end, uh, AI enabled data sciences enabled process.

I think you could see how this would be very scalable and very feasible in terms of being able to be the next generation, you know, Genentech,

Harry Glorikian: I wonder if there’ll be a big pharma that would actually have that visionary capability to bring the right pieces together. I’m not exactly holding my breath and I understand where they’re coming from.

Right. If you’re generating so much revenue, you’re sort of wait and see attitude, and I’ll just. But again, I, it always reminds me of Kodak.

Gini Deshpande: Yeah. Yes. I think that, you know, in some regards I think the. Uh, disruption, if you will, is going to happen from the biotech sector. I don’t think that this is going to be coming from within pharma.

I think, uh, part of it, as you rightly pointed out is they are under a different sort of pressure and they’re under much more immediate term. Um, objectives that they have to, to accomplish. Whereas the biotech industry has been much more disruptive. I mean, all the data points to innovation happening outside of pharma, 70% of their pipeline is actually, um, coming from the outside.

If you look at what’s, you know, all the way up to phase three, so very clear that they have embraced innovation. From the outside. So I think the stitching together, these pieces and the next sort of generation data sciences driven Genentech, like company is going to come from the outside. I think that’ll probably, um, be able to move a lot faster and be more nimble and, and, and show some early wins that, that, uh, can be done in a different context then in a more, um, structured environment that that pharma has to deal with.

Harry Glorikian: So you’re here, you know, you’re in the bay area, right? I mean, these data scientists aren’t falling off trees, right. Especially with the, with the right combination of knowledge and capability. So, uh, you know, I, I ask almost everybody, this question is how do you find them? How do you train them? How do you bring them into the tent and, you know, sort of, you know, get them to that next level, because.

You’ve got Facebook, Google, and everybody else down the street. But how do you guys at NuMedii approach this very core asset that you need?

Gini Deshpande: You know, that’s a great question. And one that I think, you know, because we are so mission driven, it is about your work actually turning into therapeutics that could then benefit patients, that mission, that objective of their data sciences capabilities.

Coming in to really help patients is very attractive to people. So we’ve been very fortunate to have data scientists on the team that have that mindset and the capabilities that are acquired and our data scientists are actually cross strain. So we have individuals who have either a PhD or an MD PhD that.

Really can understand, uh, both the biology side of things, as well as the computational aspects of things. Uh, and that becomes really critical in this particular field, because if you don’t understand that, you know, someone who does, um, who writes algorithms at, at, at Facebook, isn’t going to appreciate the nuances and the challenges of biological data.

You really need people who understand the complexities of what they’re dealing with and the nuances in this data. Um, the other is, you know, we looked at this holistically in terms of talent and recruiting talent. And one of the things we were very focused on early on was ensuring that we had a blended team that was not just data scientists, but also drug developers, because one without the other is sort of incomplete.

Right. And you really need people who have. Been there done that in terms of drug discovery and development and understand all of the challenges in taking what’s coming out of the technology and then turning it into a viable. Um, drug candidate that can then get into patients. Those processes are not trivial.

So the technology is sort of step one, if you will, of this long process, um, the components that are required to then test your drug and in vitro assays. So testing them in cell lines, testing them in animal models, those require a level of expertise and, um, A depth of understanding of, of disease biology and how far these models are relevant for advancing the molecule forward, that you really need people who are trained and steeped in that discipline.

And we’re very fortunate to have people on our team who spent 15, 20 years at pharma doing a lot of this work. So they collaborate really closely with our data scientists and that’s what it has enabled us to be successful. We have identified a novel target for IPF that, um, impacts three different, uh, disease processes, uh, at the same time.

And being able to then test that desk compounds against that target in, in vitro and in vivo assays has been possible because we have the right skill sets that are on the table that can look at this and go, okay. Sure sort of the models that we’ve got to test this in, and here’s sort of where the evidence is going to be helpful because you’re using that to gate the, the development of your, of your asset.

And so if those gating factors aren’t applied correctly, the you could have the best technology in the world, but that will not result in a compound actually making it to the clinic. And so my view has always been that you want to have people who have expertise in, in diverse domains, all around the table, interacting very closely.

Harry Glorikian: So I mean, I always think is that. I always think that’s either on the job training, right. That they’ve gotten somewhere and you’re pulling them in, or we need to rethink how we, we educate these, these in, in a college setting or in university settings so that they. Have a data analytics understanding that gets incorporated into their biological sciences.

Gini Deshpande: So we’re fortunate that there are a number of training programs that have been instituted all around the country. Now where you have training programs specifically focused on, uh, biological data sciences that are focused on both the computational and the biology aspects of, of things. And so there are training programs that are.

You know, produce some really phenomenal graduate students. Uh, who’ve gone on to do really well and have either gone on to become faculty members or have, um, taken up jobs in industry. So we’re fortunate to have people that are, that are getting trained in these sorts of programs, but you’re right in that you do need people who have.

Uh, experience within the pharmaceutical industry.

Harry Glorikian: Right

Gini Deshpande: Uh, and so we actually end up having people who might be, you know, out of a training program and not have a lot of industry expertise or experience that then get mentored by the people who do have the industry experience. Um, so it is a little bit of having to take, you know, really bright, really motivated individuals.

And then mentor them and nurture them and train them a little bit, uh, in terms of understanding the discipline and the rigor that’s required for drug development, which is very different from data sciences, um, and, and the work being done in academia. They’re, they’re completely different in terms of the mindsets and the rigor that’s that’s utilized, but it’s great when you have the right mix of people because they learn from each other and it’s, that’s, what’s makes for a fun, productive, collaborative environment.

Harry Glorikian: And so as being near, you know, say Stanford or, UCSF or any of these institutions helpful in that way of trying to identify these people that understand

Gini Deshpande: To some extent? Yes, because there are, you know, seminars and other things and other venues where you can network with and, and, and recruit people like Bappa, we’re also fortunate that we can recruit from, uh, the global talent pool.

We’ve got individuals who are. Uh, who’ve come to us from China have come to us from other countries. And, and, you know, we take the best of, of people, uh, no matter where they may be living. And, and because of our ability to, uh, work with data sciences, they’re not restricted or tied to a geographic region either.

So it enables them to have the flexibility to work in different locations. So that’s, that’s attractive for them and beneficial for us.

Harry Glorikian: Yeah. Yeah. I guess, I mean, thinking about this, you know, if you want to talk about to a certain degree remote work, right? Uh, on the data scientist side, you don’t. All have to be sitting around the same table necessarily.

Although when you do, if you’re going to do any testing, you’ve got to have a central location, but everything up to that point allows you to sort of have the best and the brightest from different locations.

Gini Deshpande: Absolutely. And so we have a core team that’s here all the time and we tap into talent pools outside of our region as, and when needed.

So it enables us to be very capital efficient and yet get the best of the expertise that we need.

Harry Glorikian: Excellent. Well, this was great. I, uh, you know, I enjoyed the session. I look forward to, uh, you know, our time at the conference together and, um, I hope we get to interact more in the future.

Gini Deshpande: Wonderful. Thank you for having me on the podcast.

I really appreciate it.

Harry Glorikian: Thank you.

Harry Glorikian: And that’s it for this special series of AI machine learning and analytics. If you enjoyed Moneyball medicine, please head over to iTunes, to subscribe, rate, and leave a review. It is greatly appreciated. Hope you join us next time until then farewell.


Related Posts