AI-Powered Precision Oncology | Aditya Pai of Genialis | The Brainiac Blueprint Podcast
- Acy Rodriguez
- Nov 9
- 36 min read
Updated: Nov 24
In this episode of The Brainiac Blueprint Podcast, we’re joined by Aditya Pai, Head of Business Development at Genialis, to explore how AI is transforming precision oncology and shaping the future of cancer care.
Aditya shares how Genialis uses machine learning to develop next-generation biomarkers, support pharma and biotech companies, and ultimately help bring life-saving medicines to patients. He also opens up about his personal connection to cancer and the book he co-authored, A Race Called Life, inspired by his mother’s journey.
Full transcript below.
🎧 Watch or listen to The Brainiac Blueprint Podcast:
Spotify: https://bit.ly/4n72kUa
Apple Connect: https://bit.ly/3VMNtlH
Youtube: https://youtu.be/KBvajJXetOs
⏱ In this episode, we discuss:
00:00 | Intro
01:18 | What Genialis does and the role of precision oncology
03:10 | A Race Called Life and a personal mission
04:54 | “I think AI is..”
06:00 | Next-generation biomarkers and the Genialis “supermodel”
10:00 | Why drug development is so long, costly, and risky
16:00 | Patient-first philosophy and unified testing vision
24:00 | The importance of data quality and global partnerships
44:00 | The future of AI in healthcare and at-home diagnostics
50:00 | Rapid-fire: workflows, aha moments, and hockey
🔗 Aditya Pai
Genialis → https://www.genialis.com/
🔗 A Race Called Life Kindle Edition
📲 Connect with Left Brain AI
Instagram: https://www.instagram.com/left.brain.ai/
X (Twitter): https://x.com/left_brain_ai
📣 Subscribe & Share
If this episode inspired you, taught you something new, or gave you a different lens on AI in healthcare, share it, leave a comment, or tag us.
Let’s help more people stay brilliant.
Know someone who would make a great guest?
Email us at podcast@leftbrainenterprises.biz
Episode Full Transcript:
Kyle: All right, welcome back to another episode of the Brainiac Blueprint, where we discuss the intersection of AI and how it impacts business and the world around us with our esteemed guests. I'm Kyle Lambert, founder of Left Brain AI and Action Hero Marketing.
Today is another episode as part of our healthcare series. We're going to be discussing AI as a core feature of precision oncology.
With that being said, today's Brainiac is Aditya Pai. Aditya, welcome to the show.
Aditya: Thank you very much. Really appreciate it. I'm delighted to be here.
Kyle: Awesome. I'm second guessing myself. Did I say your last name right? Is it Pai or is it... there's another pronunciation?
Aditya: Pai is absolutely right.
Kyle: Pai. Awesome. Nice. Hit the nail on the head there. Well, cool. So you are the Head of Business Development at an organization called Genialis. They're doing some really cool stuff and I'm excited to jump in, but tell a little bit more about your role and who Genialis is.
Aditya: Absolutely. So I head business development at Genialis. We are a precision oncology company and we help bring the most promising molecules for pharma biotech companies become life-saving medicines.
We do so with AI, machine learning, which is at the heart of our technology, and essentially help pharma biotech companies get better drugs to oncology patients, but also work with diagnostic companies in terms of bringing forward tests that are meaningful and are meant for patients who will take those drugs as well.
Kyle: Amazing. Like I said, I'm excited for this conversation. I'm going to put you on the spot here. Do you by any chance know what percentage of us are impacted, maybe it's one degree or two degrees separation in terms of knowing somebody that's had cancer?
Aditya: I would say it's very high. It would be more than 80%.
Kyle: My guess was 75. I was going to say, I feel like three out of four of us probably do.
Aditya: Yes, and touched in profound ways as well.
Kyle: Yes. Well, that's why it's so cool to hear the work that you're doing. I'm excited to learn more.
I find it very interesting because I think most people, when they hear AI, they kind of think about robots and, you know, Hollywood will have you think it's taking over the world and, you know, it's really just kind of doing tasks and things. But you guys are really leading the forefront here in terms of just health care and trying to improve our lives from a disease that everybody has been impacted by.
Before we jump in too deep into that conversation, there were two things that I wanted to bring up first. Just a slight tangent - I think it's very cool that you are an author. I think it'd be great to let everybody know you've written a book. I believe you said it was The Race Called Life. Is that right?
Aditya: Correct. Correct.
Kyle: Is there anything you want to share about that? Do a quick plug?
Aditya: Yeah, so A Race Called Life was a book inspired largely by my own personal encounter with cancer with my mom, who was diagnosed with cancer in 2012, non-small cell lung cancer. And she benefited from a targeted therapy, fortuitously. At that time, there were not that many therapies available.
And she passed away in 2015 after, in fact, having lived for a thousand days, whereas her oncologist told her she would live for no more than 100.
On her five-year anniversary, death anniversary, I just wrote a blog on LinkedIn that was around how technology has changed, how drugs have changed even in the five years since her passing. And that was picked up by a person who became a co-author on the book and said, Wouldn't it be wonderful to write a fiction story about families that are personally impacted by this, where access to genetic services, medications, targeted therapies are difficult, and bring some of the psychological, social, as well as technological dimensions?
And that's really what inspired the book. We did it on Amazon Kindle. It's available. All proceeds from that book go to Memorial Sloan Kettering. So it was not a commercial venture by any stretch, largely one meant to inspire people.
Kyle: Very cool. We'll make sure to include a link in the description of our posts and everything here.
I wanted to bring that up just because I think, again, it really helps to set the stage of how everybody's impacted. And this is a personal mission for you as well as not just a career type of thing. It is important in many ways. So thank you for sharing that.
As I mentioned to you, I also like to have all of our guests finish this prompt to really set the foundation for our conversation. So please let everybody know, I think AI is...
Aditya: Very cool, innovative, and game-changing.
Kyle: Game-changing indeed. All right. So let's set the foundation here.
Apologies, my dog is joining the conversation, shaking in the background here.
So let's build the foundation here. Again, Genialis is a precision oncology organization. I think people can make some assumptions as to what that means, but can you give kind of the one-level, one-minute explanation of what that means and how you are approaching your day to day and your treatments and everything like that?
Aditya: Absolutely. So when we say precision oncology, it's about finding drugs that work for the right patients at the right time, every time. And that's what Genialis really believes in.
We do so with a machine learning approach, but at the heart of it is understanding the biology of cancer. I think the biomarkers, as we call it, are next-generation biomarkers, and those are really different from what the conventional thinking on biomarkers is.
And so they incorporate the entire biology of the cancer, and by understanding that, we are able to work with drug companies, biotech companies, and effectively, through our machine learning technology, help bring molecules that become life-saving medicines to life.
Kyle: So let's expand on biomarkers. I personally didn't realize that there was kind of legacy, if you will, and next-gen. So can you provide some tangible examples of what they are, what it is that you're looking at, how they signal cancer or, you know, point you guys in the right direction for your treatment and your research?
Aditya: Indeed, and I think all of us come in touch with biomarkers when we visit our doctor. So a simple example would be something like a hemoglobin A1c. That's a biomarker that tells you an average of three months of your blood sugar levels, and if it's above a certain point, it indicates that you're diabetic. That's a very simplistic biomarker that's very pervasive and used frequently.
Then when we look at cancer, it's a little more complex. The more traditional biomarkers tell if you have a certain mutation in a certain gene at a certain point, and that would be a simple nucleotide change. So I'll give you an example of KRAS. KRAS is one of the most…
Almost 30% of cancers have a KRAS mutation in them - lung cancer, colorectal cancer, pancreatic cancer, in which case it's more than 90, 95% even. But in non-small cell lung cancer, if you have a mutation at the G12C point in the KRAS receptor, then that is used as a biomarker for effectively stating that you are confirmed that this gene is mutated and therefore you become eligible for a drug.
There are only two drugs that are on the market right now, many in development.
That's the simplistic way of looking at it. What we do is very different. We actually model the entire biology of KRAS. KRAS is the start of a very complicated biological pathway. And so by modeling the biology, we are able to build algorithms that in fact are representative of that biology and not just looking at a single change.
That's what we mean by newer-age, next-generation biomarkers that are more comprehensive and can really help pharma companies understand several aspects: Why does a drug work? Why does it not work? Why do some patients become resistant? How long will a patient benefit by being on the drug?
The G12C mutation can tell you that the person has cancer, non-small cell lung cancer, because of this mutation, but it doesn't tell you the durability of response by taking one of the two medications that are on the market.
And we are able to do that by using our technology. We're able to be much more precise. We're able to effectively tell when a patient will benefit from that particular drug or perhaps a combination therapy that otherwise would not be achievable.
Kyle: Interesting. So you said that you can determine when they would benefit from the drug. So, like, these algorithms that you guys have that are at the core of your decision making, it seems like it takes into account the status quo. But I assume there is kind of a future aspect of if you keep living in a certain way or, you know, of course, it takes into account your own demographics and everything.
So it factors everything in to kind of know you - at least, I guess, estimate your path or your trajectory here?
Aditya: Yes, it does. And we have presented a poster recently at ASCO where, you know, there's a small subset of patients that we looked at, but by and large, we showed how we can predict when a patient would benefit from a certain therapy. This was a KRAS therapy, and at what point the treatment would stop, and what other treatments that patient would benefit from.
Naturally, you need a large number of patient studies like that, but the technology is able to stretch its boundaries to that extent. And particularly when it comes to the complexity of cancer, if you just focus on a single mutation or a single change, that's not enough. That will give you a response in 30, 40% of the time, but you're still missing a whole lot.
So we can do a lot better, and that's what our technology relies on. And I think that's what makes us inherently different. We focus on the biology of the cancer, and I'll bring up this point maybe a little bit later in our conversation - our foundation model for cancer that really enables that to happen.
It's a flywheel of sorts that rapidly accelerates the configuring of biomarkers that can, in fact, achieve this purpose fairly rapidly for pharma biotech companies and then also help build robust diagnostic tests that are also central to who should be taking what type of drug.
Kyle: Very interesting. Very interesting. So are you able to provide any kind of numbers in terms of volume of biomarkers that is incorporated into your analysis, whether it be at one specific person or kind of the overarching analysis that is letting you believe, okay, we're on the right path, this could actually be a treatment?
I assume, again, you're using AI, so this is probably millions and millions, if not billions of data points that are being incorporated, which is something that a human cannot do. So I'm interested in the volume of biomarkers and data that you're using here.
Aditya: Of course. So this would be probably a great time to bring up our foundation model for cancer. And this is referred to, you know, as our supermodel, because it is a model of models.
And effectively, it's been modeled on hallmarks of cancer, which is a well-known fact in cancer biology. This was introduced by Hanahan and Weinberg as 14 different hallmarks that really explain why a person might get cancer.
And cancer is, of course, a very heterogeneous disease. Each cancer has its own reason. It could be a DNA damage repair issue. It could be a signal transduction issue. There are so many different reasons that a person might get cancer.
So these 14 hallmarks effectively cover that entire landscape of cancer. We trained our foundation model on those 14 and, in fact, came up with about 150 signatures. We call these biomodules, and these represent inherently, you know, entire pathways that correspond to biological processes that define specific cancers.
And these were derived from something also very unique to what we do, which is our preferred analyte for doing this is the transcriptome, which is RNA, as opposed to DNA. It's not that we don't use DNA, but we simply think that RNA offers a richness that is significant. It explains why something might be happening to a person as opposed to a more historic account of what might have happened, which is what happens with DNA.
All three are perfectly fine - DNA, RNA, or protein - but we specialize in RNA and, in fact, take the entire transcriptome, the whole transcriptome, and select features from that that go into building these biomodules.
So we're not, when we refer to our KRAS ID, which is our first-in-class biomarker for KRAS drug inhibition prediction… it can help in a number of ways. It can help a drug company working on a KRAS inhibitor to know why the drug is working, why there might be resistance. Can it be used as a predictive biomarker? Can it predict patients who will benefit from the drug and for how long, and time to treatment?
And that's really the inherent basis of that theory. And so the foundation model for cancer has these 150 biomodules that - think of them as Lego blocks - that can be configured in many different ways for any drug target in the cancer space.
So whether you're dealing with a DNA damage repair agent, or you're dealing with an antibody–drug conjugate, or a KRAS inhibitor, or a checkpoint inhibitor, we have the underpinnings of a biomarker already that we can configure and effectively model and configure for a specific drug that belongs to a particular drug class.
Kyle: Interesting. Lots going on there. It's very interesting to see how it's all approached and how it all comes together.
So I'm going to ask a question that may honestly conflict with the whole purpose of precision oncology, but I think it kind of outlines that there is some complications to this, but you obviously have a North Star.
So if we're thinking about your typical person - so again, that's where I think it's probably far from what you guys are trying to accomplish. You don't want to have a typical process or anything like that.
But your typical cancer patient that is using a treatment maybe that Genialis helped to develop - what does that process look like? And specifically, I'm asking about, you know, there's preventative, there's ongoing, there's helping make sure that a relapse doesn't happen, things like that.
So is there constant tests that are happening and then there's new biomarkers that are reviewed and you're hoping to again hit certain milestones? What does that, again, for lack of a better word, typical kind of process look like?
Aditya: Yeah, so certainly I think with everything that we do, one of our core values is people first. And that pertains to not only people within our company in terms of the well-being of every individual in our company and the work–life balance, but also people first when it comes to patients and finding, therefore, the right drug for the right patient.
Our long-term goal is to effectively create such a test that is put in the hands of a physician that can order it and has a reimbursement path, because payers are a big part of it, and, in fact, can give complete information for that patient.
We haven't done that yet. What we have done is work with pharma biotech companies and partnered with a number of diagnostic companies to make that reality happen.
So for a particular typical kind of scenario, we would be working with a pharma company usually starting in early preclinical development before a new drug, investigational new drug (IND) filing is made, typically. Sometimes phase one. But typically if it's phase two and three, they've already set their biomarker strategies or how they want to run their clinical trials.
That phase one can be, or earlier, can be very informative because it helps basically provide the drug company with an understanding, a better understanding of the drug based on our supermodel.
Our supermodel is trained on hundreds and thousands of whole-transcriptomic records from a very diverse set of data, global data. And therefore, it's already a trained model, and we effectively configure a biomarker on the basis of that.
And that biomarker can then be used after an understanding of how the drug works - we call it a mechanistic understanding of response and resistance. It could be used as a patient selection tool, right?
Which is you create a clinical trial assay, and that clinical trial assay is typically developed by a diagnostic company. And you then incorporate it for prospective clinical trials. And we've started doing that with some drug companies as well.
Now you're selecting those patients that you know, on the basis of the biomarker, will in fact respond to your drug. You select those.
And eventually, if it continues along that path, that could become a companion diagnostic as well, which is every time the drug is given to a patient, you give it on the basis of a diagnostic test that says that, in fact, they have characteristics that will enable them to benefit from the drug.
So it's really all along that clinical development path from preclinical all the way to FDA approval that our biomarkers can play a role.
And if you think about it, it's about giving that patient the best chance of durability of response, right? Eventually, that's the goal - that that patient we know will benefit from a response, or might benefit from not just the monotherapy, that specific drug, but a combination therapy as well.
So both those aspects come. So that's really what would be our North Star, which is we always keep the patient first in mind.
And that is a long process. It's a 10–15-year process that costs $2 billion. And there's a lot of economic sense for drug companies to, in fact, use this type of method because ultimately you are ensuring those patients stay on your drug for a long time. They benefit from the features of your drug that make it unique and, in fact, alleviate some of the complications of cancer.
Kyle: Truly amazing stuff.
You kind of took what my follow-up question there was going to be, and I was going to ask about how long it usually takes to get to a clinical trial or even to a spot where it's like, okay, we're good to go. And you mentioned 10 to 15 years.
Is that the normal case? Is there red tape, I assume, that kind of slows things down, or is there just, you know, you guys want to make sure you're very sure and you hit, again, certain milestones? Just if you can elaborate a little bit on that.
Aditya: Absolutely. So this is a 10 to 15-year process going from you nominating a target that you want to use - so there's a drug target discovery process even before that - to you nominating a target, to in fact then you going through preclinical work, which typically is an animal model, cell lines, organoids, and then you are ready to take it to the clinic.
At that point, you have dose escalation studies as well that happen. So you want to find out at what dose will this treatment work the best. And there's been a lot of change even in the FDA recently around some guidelines and roadmaps around a specific project, Project Optimus, that looks at optimal dose as opposed to maximal tolerable dose.
And then beyond that, it goes into phase two, three, and progressively as you get to phase two, three, not only are you enrolling more patients - and finding those patients is another challenge - but also having costs balloon.
And so having a phase 2B or 3 failure can literally wreck your company. If you're a small biotech, it absolutely can. If you're a large pharma, possibly you can withstand. But that is not only disappointing to the patients who took part in the trial with high hopes, but also to the 10, 15 years that go into that and the tremendous amount of financial capital that has gone into that process.
90 to 95% of drugs fail in cancer. And so this also brings credence to the fact why a biomarker-led strategy - the earlier it is, the better - because you're able to, in fact, learn things about your compound as early as possible.
And so if you start at phase 1B, you might have missed even some critical aspects of dose escalation or the first inpatient dosing that might have altered your strategy at an earlier stage versus investing a lot more money at that point.
Some of the most effective drugs - so I'm talking about all the EGFR inhibitors, the first, second, third line inhibitors that came up. These were EGFR mutations. My mom had one of those and remarkably successful, right? These are successful, but these were biomarkers, right? Simple biomarkers, mutations in EGFR genes.
Immune checkpoint inhibitors - the biggest blockbuster drug in the world, Keytruda - has benefited from a biomarker around PD-L1. And if you have an excess amount of PD-L1, then it tells you that you might be a candidate for this. And then its tissue-agnostic nature - so it's not treating the cancer in the one location in which it occurred, but it can be used across all tissues.
But this is also biomarker-led. They might not be the best biomarkers, and we see there's tremendous room for improvement in all of those.
But certainly it's a proof point that a biomarker-driven approach can lead to very effective medications as well. We just think we can do much better in that area with our novel approach.
Kyle: Sure. Very, very cool.
Your point earlier there about how it can wreck the company is so interesting and so crazy. As an entrepreneur, I'm self-funded to kind of start Action Hero and Left Brain.
So small business - there's always anxiety and kind of fight-or-flight mode, you know. To have someone else's billion-dollar investment and someone else's lives kind of in balance there, the stakes are so much higher. So that's so intense to think about.
I assume because of that, there has to be - and I think this will lead nicely into what your role is in terms of partnering up with the right data and everything - so you have to make sure your algorithm and your data is working right.
So what does that look like? You're relying on these algorithms. How do you know that the output is what it's intended to be and that it's accurate and that you're not just going down the wrong path here and setting yourself up for failure?
Aditya: That's a great question. So it is all about the quality of data. And data that comes from a transcriptome from different locations has to be really of high quality. And that quality also means normalizing that data so that you're comparing apples to apples - that's the analogy I can give you.
And so we have a very elegant software that, in fact, does that data normalization. Years were spent developing a very slick software that sits at the start of our biomarker development journey. And that is, you know, you get a raw sequencing file - it's called a FASTQ file - it goes through this normalization process so that what we call machine-learning-ready data…
So that data has now passed through all the quality checks after onboarding and is at a level that is acceptable internally for use.
Now, this data comes from different cancers from different parts of the world. Some of these are through our data partnerships that we have with different organizations across the world. And in other cases, it's publicly available data that is available to everybody for use.
So all of those come into play, and all of those go into the building and the training of our algorithm that has to pass certain tests.
Once that is done, we know that it's an internally validated biomarker that can be now used to test against a client’s data. That could be data coming from cell lines. It could be data coming from a mouse PDX - a patient-derived xenograft. It could be an organoid. It could be human data, right?
But at that point, we are able to take that as a baseline and apply it to the client's data.
Kyle: So I assume some of this might be your secret sauce and proprietary, so, you know, don't share anything that you can't, obviously.
But can you elaborate on these tests? You know, like you said, you upload it to a software and it has to hit certain quality tests. Again, I'm just trying to understand how you understand - is it like, okay, this hit 90% of the points that we're looking for? What does that look like?
Aditya: Correct. Something like that. So the software is intended to, in fact, do the normalization. And that normalization means that you're able to create this data that is translatable.
There are internal tests that are done. When we talk about our supermodel, it falls under this category of explainable AI. So there's no magic black box that something goes into and, you know, gives out a biomarker.
We can go down to each biomodule and show exactly how a sample is performing. And that is basically a bioinformatics, computational biology task, right? So those are technical benchmarks that are there that, again, are very evident for everybody to see when we engage with our collaborators and clients as well.
Kyle: Interesting. Okay.
So you are the Head of Business Development. You're going out trying to find this data. How many providers are there? This seems very specialized. So from an outsider's perspective, I can't envision there being that many, but I assume there at least are that can provide this information.
So how many are there and what does that search look like for you? What does that relationship look like for you? Kind of tell us how you're developing business and helping Genialis.
Aditya: Of course. So I'll classify it into two. One is data partnerships, which is we look for data partners that have a like-minded approach, very collaborative, interested in doing good. And so those collaborators are from different parts of the world.
I can give you a couple of examples. Some are very specialized. The Pancreatic Cancer Action Network would be one. These are all publicly known. And so we work with them, and they specialize in pancreatic cancer. A fantastic organization, very, very collaborative.
A second would be in the Middle East. We work with Sidra, and they provide a very unique type of data set for us to use as well. We have collaborations with Academia Sinica in Taiwan. So these are just examples of how we seek out data partners.
We meet them at conferences, we learn about them through publications, and generally there's data there that is of interest - that they have transcriptomic data, but also have real-world evidence as well.
We've also talked about - you've probably seen our collaboration with Tempus as well, Tempus AI, and that's another one of those where our KRAS ID algorithm was trained on our data, but we needed clinical validation of that and rapid clinical validation.
So Tempus was able to provide us with real-world evidence on that - specifically, non-small cell lung cancer G12C-mutated patients who were treated with sotorasib, on whom we had pretreatment RNA sequencing data. So it gets pretty specific, but, you know, the devil is in the details.
And that's just one example of how we collaborate with our different partners. And that is really on the data partnership side. That allows us to continue to build our database so that we can train our algorithms and, of course, use them with different pharma partners.
The other part of my role is actively seeking out drug developers - pharma, biotech companies, diagnostic companies - and these fall again into two categories. The drug development business, you have biotechs and you have large pharma, and they typically have programs that we follow.
These might be areas that are difficult, like DNA damage repair or KRAS, or areas where we have done a body of work. Again, the aforementioned areas, but also the tumor microenvironment with checkpoint inhibitors and angiogenesis.
And so the business development effort is around: How can we help you bring drugs to market with a biomarker-driven strategy that will provide even more success for your drug?
And that's a very collaborative process that goes on at every different stage of drug development.
With diagnostic companies, you know, it's about - at some point, you have to convert that biomarker into an assay. And we are what we call a dry lab. We work with sequencing partners, labs, to provide us with sequencing data, but also the ability to follow regulatory guidelines that are very stringent in developing a clinical trial assay that can be used for prospective clinical trial use.
It could be a companion diagnostic. And so that's where partnering with diagnostic companies comes into play.
We also think about eventually building tests that these diagnostic companies will bring to market. And there's complexity there because there's a payer component involved. There's a prospective clinical analytical validation method also in play here.
So these are, again, very important partners. So again, to summarize, it's the data partnerships aspect, which is about bringing data in, normalizing and making sure the data is high quality from a variety of heterogeneous cancers from different populations in the world.
And then, of course, a second aspect being working with pharma–biotech with their drug programs, and then of course, with diagnostic companies to either bring a clinical test to the market or clinical trial assay/companion diagnostic space.
Kyle: Kind of reminds me of almost a mill. You're getting your raw material, it's coming in, you guys are processing it and then, you said, normalizing it. And then you have an output that can be actually used by humans or organizations in a lot of different ways. So it's cool to kind of see the whole process come together like that.
Aditya: Yes, you're absolutely right. And this is what I refer to as the flywheel at the start of our conversation, which is our Genialis supermodel, our foundation model for cancer.
It is a flywheel that, with each data piece coming in - it's literally like having two gears on a bike - that you're able to bring in this data and every piece of data that comes in, some of the signatures that we've built become stronger and more robust.
Kyle: Very interesting.
So in your previous response, you had talked about tests. And I want to jump back to that - let's not forget that. But before we do, I want to wrap up the conversation about bringing in the data and making sure that it is inclusive and global - bringing in all kinds of demographics and touching all the geographies and even socioeconomics and everything like that to make sure, again, the full picture is painted.
So you are an experienced business development rep, we'll call it - leader. You're very experienced in this space.
What is just an automatic red flag or makes your spidey senses start to go off, be like, okay, we need to be a little careful here, this might not be the right partner for us in either direction, whether it's providing the data or the output management, if you will?
Aditya: Yeah, so I think so much of this is about working with your collaborators and educating them. It's not a trivial process. When I say business development, it's very, very consultative.
There's a lot of education involved because, you know, typically when you think about a biomarker, you think about a mutation. You might not think about the entire complicated KRAS–MAP kinase pathway that is complex - it has various genes involved at every stage in it.
And so there's an educational component. Then there's a component of showing you how this feature reduction happens, how you start with the entire transcriptome and how you distill it down to specific features that define the biology of that cancer, that become these discrete biomodules that can be 150 of those that can be used as Lego blocks as you choose.
So there's that educational component. Typically, red flags are when a client might not understand that. And so we have to take the time to always say that if you can't explain your technology, it's not the client's fault. It's up to you to explain it in a way that enables them to understand.
So we make sure we do that. That's number one - is try and explain this in several ways so that they understand what is it that we're trying to do, the sophistication of what we're trying to do.
The second part is the data. So the data is entirely dependent on who created that experiment at the client. So if you decided to just get RNA sequencing done and you didn't really care about the number of total reads of RNA that you were looking at, the sequencing might be too shallow and the client might not even know that until you get the data.
That could be one such factor. Another factor could be just the way your experiments were designed, right? And so we've learned to be very diligent about that, right? About looking at certain data characteristics.
And you learn as you go along as to: Why are you giving us this type of data? Why are these samples performing in these ways? And so it's a very, very consultative process with the client.
And it's truly collaborative, right? Because you're trying to use this technology to find value, right? Give them value in explaining something that they didn't know about their drug - or knew, but that really was corroborated by what we found.
So I would phrase it in red flags as really more client education and bringing our experience to bear, just trying to explain to them the importance of good data quality.
Kyle: Absolutely. So there are standards that they need to adhere to. I assume there are industry standards as well as - does Genialis have their own ones, that they take these industry ones and they kind of maybe even make it even more tighter or adapt their own use case?
Aditya: Yeah, I mean, we suggest - like, if somebody asked us, What kind of sequencing depth would you like? we can provide that, right? Absolutely.
Sometimes we don't have the luxury of that because the data has already been sequenced, right? And then we have to look at that - look at what data we get.
Obviously, a luxury would be to influence that from the beginning. If somebody asked us, okay, in an ideal world, if you could dream it, what is the type of data you're looking for, we would be able to tell them exactly what type of data.
Kyle: Got it. Awesome. Very cool.
So, all right, I want to circle back. We talked about tests when you and I originally connected. We had a discussion about diagnostics and innovating what that process might look like.
So as I understand it from our conversation, currently there are multiple tests that are developed, conducted to understand even different variations of, like, lung cancer or things like that. And it can be, to your original point, tough on the patient. You always want to try to keep the patient in mind and make it as easy for them.
So can you - I think your vision was having a single comprehensive test, is that right? So can you kind of elaborate why, you know, why we're not there right now and where you think that you can go with some of these processes that you have in place here?
Aditya: Yeah, so I think this is all an evolution of the technology and the science and our ability to package this in a way that is understandable by a physician ordering the test, is understandable by a patient who's going to benefit from that test, and then also with a very clear value-driven reimbursement path that allows for the payment of that test.
So these are the three components in this. And so I think - I'd written an article on LinkedIn about lung cancer and how, you know, it used to be just a few mutations in the EGFR gene or a fusion called an ALK fusion.
And then as more and more drugs came to being, they started becoming even more focused. They started becoming focused on those patients that were resistant to first-line therapies, who had a certain mutation, who could benefit from second-line therapies.
And so how do you now start navigating 10 to 15 different drugs, each with their own mutation profile? How do you know what to order, let alone how to interpret?
And while you have comprehensive DNA panels, RNA panels that are offered by several diagnostic companies, I think a case can be made that ultimately you can use all this information to your advantage and have one test.
You reduce the friction across all those levels that I mentioned. And I refer to this word friction because that's exactly what it is. It's across payers, providers, and importantly, patients.
And so if there was a single test - and this is aspirational for lung cancer - and that accounted for all of that, wouldn't that be great, right?
But the dynamics of how each of these evolve is: some of these tests are linked to specific mutations, to specific drugs. And so it's not as easy as saying, Let's just have one test, but it is absolutely an aspirational piece that I think we will eventually see happen - that there will be more of a consolidation of, How do we make this easier for all the stakeholders that are involved?
Kyle: Sure.
And you kind of brought up another question there. So whether it be this unified testing approach or even just a pharmaceutical treatment that Genialis brings to market, you are in Canada. I'm in the United States. We obviously have very different healthcare systems, as does many different parts of the world.
There's always, again, different levels of accessibility and things like that. So I'm curious how you think about that. If you think, again, maybe it's this unified test or any other treatment, how different… Is there kind of a process that you think about of trying to make it more accessible to people?
Or is it, you know, we go to the United States where we know we can get buy-in, and then we'll have data and other countries will follow suit? You know, what does that look like for you?
Aditya: Yeah, so I think it depends. That's a great question because I think it really entirely depends on each country's healthcare system. Canada is very different. And, in fact, Canada has - you know, each province has its own healthcare system.
And so a drug that is approved in Ontario might not be approved in British Columbia, for example.
And the type of tests that are ordered through typical cancer labs in the province are very specific, as opposed to the commercial tests that you can go to a commercial diagnostic provider in the U.S. and have that done out of pocket as well.
It's very different in, say, a different country - an African country, or in Southeast Asia, or South Asia, right?
I think central to that also is access. So, okay, so you have a test, and you do this test, and it tells you, Here's this drug that you as a cancer patient would be eligible for, but now the drug costs $20,000 a month. So how do you now navigate that?
So there's inherent complexity also in access. Making a test accessible doesn't mean the drug is accessible, right?
So I think that's a really important aspect that drug companies would have to work with diagnostic providers in different jurisdictions because what's true in one part of the world might be entirely different.
We even see that across the Canadian and U.S. border, where things that happen in Canada are very different, right? In terms of how the regulatory path they go through, the market access path they go through, the payer component - very, very different.
And I think you're going to see a continued evolution of this.
Kyle: 100%. It's such a complicated situation that changes almost weekly, if you will. So it's interesting to see how - it will be interesting to see how it all pans out.
I think, again, the data, the proof is in the pudding, right? When people start to see things work and it's like, okay, we need to figure out how to make this work for our, you know, our patients and our citizens and things. So it'll certainly be interesting.
Great. So the last thing - I want to pivot quickly a little bit to AI and think about the patients.
When you and I were connecting, we kind of talked about the next frontier and where you think AI is going to be plugging in a little bit more in terms of it being in the hands of the patients. And you had some interesting ideas.
So I wanted to kind of see if you could elaborate on how you think AI is going to be helping the end user a little bit here.
Aditya: Of course. And I think the continued evolution of generative AI is central to this, as is the concept of agentic AI or co-pilots, right?
And information is power, and information that is easy to access is very different. So think about even a few years back: you would go into your Google browser and search for molecular tests for a particular cancer, and you would get, you know, a list of different options and you would have to click through each one.
Now you can put a very smart query, and you don't need to really be, you know, one of those query experts either. Anybody can put a query into a generative AI - Gemini, ChatGPT system - and get a very comprehensive view of what kind of tests they would be eligible for.
They can feed their cancer report and get recommendations on the basis of that. Now, are they right? You know, again, you have to be careful how that information is used, but information is power, right?
And I can make the analogy to, you know, when I used to work as a genetic counselor, you know, back before I started my business role. This was, you know, basically where the internet had just become pervasive.
And the pre-internet time, when I started practicing as a genetic counselor, I was really the information guardian. That changed to people coming to my office with a stack of reports saying, This is what the internet told me.
Whereas now, that has taken a completely different level of sophistication. Now you actually have a far more refined…
And so, where does the information lie? The information lies in the hands of those who can access it easily and make decisions very easily on that.
So I do think that the whole concept of co-pilots, of agentic AI, will serve, I think, a very important purpose. And I think that purpose is going to be more educational than anything else.
For sure. I mean, if I think about where some of the biggest gaps are in all this technology that is outpouring - How do I actually use it? Right?
How do I use molecular technology, whether it's transcriptome or a panel or a single gene test or, you know, this concept of precision oncology - what is precision oncology, right?
And I'm just talking about cancer, right? This is one area. This extends through any other therapeutic area as well.
So I think there's going to be a significant evolution in this. I think the FDA also has kind of provided a roadmap which looks at the use of AI. And this can be to model different cancers, the progression of those cancers, the use of AI to, you know, essentially pick out synthetic control arms, or use organoids in a way that you are able to get to an IND faster.
So in all those aspects, I think AI is going to continue to see a massive…
So you asked me to complete the sentence, and I said game-changing. That's what I mean by game-changing.
Kyle: 100%. I think if you even just look at ChatGPT-1 to ChatGPT-5, it's a whole different ball game.
Like you said, the agents are available. It's a compounding effect. It's all building and just accelerating even faster.
So I 100% agree. I think that these tools are just going to become way more entrenched in our day-to-day in many facets, healthcare included, and it's going to be interesting to see.
One of the original things that I thought about that AI, I think from a healthcare perspective, is going to open up is - I immediately go to like these 23andMes with the mouth swabs or maybe even like a smart toilet or something like that.
Do you foresee it bringing tests into the home, the at-home tests and things like that even more so?
Aditya: Why not, right? And so if you think about cancer, the best chance of you escaping cancer is not getting it in the first place - your lifestyle and habits.
Second is early detection. That early detection is important because you surgically remove the tumor from the area. And if you've gotten all of it out, then theoretically you're cancer-free and you're in remission.
But the more you can pay attention to those kinds of signals and incorporating some of the age-related even benchmarks that are out there - colonoscopy at this age, mammogram at this age, Pap smear at that age - you know, all of these, I think, are important.
But there could be, in the future, a toilet-bowl test that gives you an early detection of, you know, gynecological or urological-type cancer very early on.
And the early kind of advent of that test is fecal occult blood testing, and that's done through a kit that you get sent - in Canada, at least - by the ministry. And you do that test and you send it back, and then you get a report back saying positive or negative, right?
But this is the next dimension in that test where there's something in your toilet that is sending something back to your phone, probably, telling you you're at risk for this, this, this - and what is the percentage of your risk. And that is all calculated through predictive algorithms in the background, right?
Again, it's hypothetical, but did anybody think you would have ChatGPT three years ago or four years ago? No, right?
Kyle: I'm now picturing people being afraid to go poop so that they don't get bad messages on their phone.
Amazing stuff. Amazing stuff.
So Adi, I wanted to do some rapid fire questions real quick, if you don't mind.
So first, if you could click your heels three times and have just this automated workflow built into your day-to-day with no cost or restrictions, what would it be?
Aditya: It would be to get the best drug to the best patient at the right time.
Kyle: All right, fair enough. That's a complicated one right there. I wish, right?
What is your favorite or most memorable aha, light-bulb moment that you've had over your career?
Aditya: I think it was when I saw the power of immune checkpoint inhibitors. And the first time I realized that, wow, you're not just looking at a particular cancer localized to your particular organ. Now you're looking at a tissue-agnostic use of a checkpoint inhibitor.
That is a pretty big revelation. And second would be just in university, understanding how molecular biology works - the fascination with the complexity. And the complexity is why we are where we are today. It's not simple, right?
Kyle: Oh, yeah. Very cool.
If AI was actually all-knowing and could answer any question for humanity, what would you ask it?
Aditya: At what point would we, with the current piece of research that's going on, at what point would we get to an 80% or greater response rate with drugs for particular cancers?
Kyle: Okay, very cool.
Aditya: Because that would be a significant number, right? Over 80% response for a particular cancer with a cure rate of that would be great.
Kyle: I think a lot of people would be happy with 80%, for sure.
We talked about you writing a book before. If you were to write an autobiography, what would the title be?
Aditya: It would probably be - the autobiography title, that's a good one. That's a good one. You've stumped me on that one.
Probably The Race Is Ongoing.
Kyle: Okay, I like that. The Race Is On. I like that. Staying on theme.
Last one - you and I are both very big hockey players, as we can - or hockey fans, as we can tell.
You were saying that you're a Canadiens fan, right? So who is your favorite hockey player of all time and your most hated hockey player of all time?
Aditya: My favorite hockey player is a Finnish player by the name of Teemu Selänne, and he was known as The Finnish Flash and absolutely wonderful person, but also a phenomenal player.
Kyle: I have very clear memories growing up with my dad trying to open up, I think it was Topps cards, and hoping to get a Selänne rookie card. So I know very clearly.
Aditya: And it's interesting because I'd gone for the big hockey tournament in Montreal - the Four Nations Cup - and I was just walking on the street and looked in this restaurant and Teemu Selänne was standing there.
So I went and had a nice chat with him.
Kyle: You did? Is that what you said?
Aditya: Yes.
Kyle: Oh, wow. That's awesome. That's pretty cool.
Aditya: I think there's a lot of - I think you mentioned hockey, but I think it extends to just my fascination with sports. Cricket is another one.
And I say that because I think there's a lot of learnings, right, about, you know, these immensely gifted athletes who all have to kind of come together and work as a team. And so there are a lot of team lessons and adversity they face in their professional games and how they overcome those.
So a lot of hard work that goes into it, but I draw a lot of inspiration from that because I'm not a professional hockey player, but I think you can apply those lessons in your own profession as well - about teamwork and tenacity and dealing with obstacles and also celebrating the wins.
Kyle: I'm a retired collegiate lacrosse player, and I still like to say that I'm an athlete. So I certainly know exactly what you're saying. There's a lot of lessons to be learned that apply to life and business and everything, you know, learning how to all march together. So incredible.
So the last part here, Adi, I want to just give you a moment for open forum. Is there anything that you're passionate about or you wanted to share that maybe we didn't touch on that you find important that the audience hears about?
Aditya: I think we've covered a lot of ground, but I truly feel that the role I'm in is about really helping patients in the end. To me, it's always been about the patients. And all this technology is fantastic, but then you're still not there yet, right?
So I have anecdotes and examples of friends who at very early ages passed away from cancer. One of my close friends is currently going through that right now, same age as me, and you go, Why should that happen?
It just means that we are not there yet, right? And you should be able to, you know, give a person who has cancer a good quality of life for more than six, eight months - in some cases even lesser - and detect it early.
A lot of great work going on with early detection work with a number of companies with blood-based tests, urine-based tests.
So I think we should all - I mean, that would be my call to action, you know, to anybody listening to this podcast, anybody in the industry - is: How do we drive ourselves to becoming better and continue this pursuit, right?
And yeah, there's a business component to it. But I think, back to your original question, How many people do you think you know who have been affected with cancer? What's the percentage? There's a common driver.
And this is just - we're just talking about cancer, right? If you can turn it into a chronic disease that can be treated regularly and still maintain a good quality of life, I think that's there.
But, you know, that's just one of several other therapeutic areas that I think we all can benefit from working together on.
Kyle: Well, you don't want me working on any tests or anything like that, but if there's anything I can do from a marketing standpoint out there, I'll get the names out there for all these organizations. So I'll do my part.
Aditya: I definitely appreciate that, Kyle.
Kyle: Amazing. Well, Aditya, thank you so much for your time. I know people can find you on LinkedIn. Is there any other channels or anything like that you'd like to share or have people reach out to for you?
Aditya: Yeah, LinkedIn is the easiest channel. Obviously my Genialis email, aditya.pai@genialis.com, that's great too. And I'd be delighted for anybody who wants to talk more about this or has an interest in any type of collaboration as well.
Kyle: Awesome. Perfect. And if anybody wants some more information on Genialis, you can find them at genialis.com. Lots of great stuff on there.
So Aditya, again, thank you so much for joining me on the Brainiac Blueprint. You've shared some amazing stuff. I'm inspired. Hopefully everybody else found it interesting.
If you do me a favor, please look at the camera and say, Stay brilliant, Brainiacs.
Aditya: Stay brilliant, Brainiacs.
Kyle: Awesome. Thank you, Aditya.
Aditya: Thank you. Thank you very much.
Comments