Transforming Cancer Care in India with AI

Daniel Levine (00:00)

The Life Sciences DNA podcast is sponsoredby Agilisium Labs, a collaborative space where Agilisium works with its clientsto co-develop and incubate POCs, products, and solutions. To learn howAgilisium Labs can use the power of its generative AI for life sciencesanalytics, visit them at labs.agilisium.com.

Sri, you're in India and doing a couple ofepisodes from there. Before we talk about today's episode, perhaps you canstart with a little perspective for our audience. Where does India sit in thecontext of the convergence of the life sciences with AI?

Nagaraja Srivatsan (00:42)

Danny, it's been really exciting. India isreally a virgin here in both AI and healthcare. I'm really excited as I go cityby city. I've been seeing quite a lot of applications of AI in healthcare.There's lots of startups, lots of people who are actually implementing AI, so it's really a great time and it's wonderful to see that kind of enthusiasmhere.

Daniel Levine (01:04)

So you'll be interviewing Vijaylakshmi  Ramshankar. for audience members not familiar with her, who is she?

Nagaraja Srivatsan (01:12)

Danny, here today I have Vijaylakshmi,Professor of Cancer Biology and Molecular Diagnostics at the Cancer Center.She's a doctorate in molecular oncology and has completed a postdoctoral fromseveral places, including the Cancer Institute, as well as Indian Institute ofScience and Columbia University. She has over 25 years of experience inmolecular oncology. Her primary focus has been in tobacco related cancers likeoral, lung, and human papillomavirus, and her objective is to translateoncology research to patients from a bedside perspective to both help them froma prognostic as well as a predictive biomarkers perspective.

Daniel Levine (01:53)

What is the Cancer Institute?

Nagaraja Srivatsan (01:54)

The Cancer Institute was actually established in 1954 as a non profit organization. It's a regional center forcancer research and treatment and recognized as the state's cancer institute.It's a pioneer in cancer care and offers state-of-the-art treatment for severalcancers.

Daniel Levine (02:11)

And what are you hoping to hear fromtoday's interview?

Nagaraja Srivatsan (02:14)

I'm really excited because Vijaylakshmi isin the nexus between AI and molecular diagnostics and clinical care. So it'sgoing to be a very exciting conversation on how you can use AI to solve forwhat we call current world problems, but also future diagnostic problems.

Daniel Levine (02:33)

Before we begin, I want to remind our listenersthat if they want to keep up with the latest episodes of Life Sciences DNA,they should hit the subscribe button. If you're enjoying the show or want tosuggest topics you'd like us to explore in future episodes, post a commentbelow. And if you'd like to enjoy the show on the go, you can find it on mostmajor podcast platforms. With that, let's welcome Vijaylakshmi to the show.

Nagaraja Srivatsan (03:01)

It's so great to have you here today. WhatI wanted to explore today is really the nexus between what you're doing withcancer research with the new tools coming up with AI, and how that can beapplied to a marketplace like India.

Vijaylakshmi Ramshankar (03:17)

So most of the cancer research programsthat we have undertaken in the department is pertaining to the most commoncancers that we find in India. Like for example, breast cancer, which is aboutone fourth of all the female cancers is breast cancer in this country. Cervicalcancer, is the, you know, we have the highest global incidence of cervicalcancer. And oral cancer is again due to the extensive tobacco use in thiscountry. We have a lot of oral cancer.

We are actually the oral cancer capital ofthe world. Wow. Yes. And then because of a lot of environmental pollutions thatis there across, we have a lot of lung cancer these days. The lung cancerincidence is really increasing and colorectal cancer. So these are some of thecommon cancers which are there in the country. And the department research iscurrently catered towards these cancers. Except the colorectal cancer, we aredoing research in the cancers that I mentioned.

Along with that we are also working onleukemias which are you know very rampant childhood cancer which is there inthis country. So, this is the research interest of the department.

Nagaraja Srivatsan (04:25)

That's a quite a wide variety of cancerresearch to be done. So how do you start? You know in your moleculardiagnostics, is it post diagnosis or are there measures to be more preventiveand...

Vijaylakshmi Ramshankar (04:38)

It is both. So when it comes to moleculardiagnostics, when it is offered to the cancer patients, it is after thediagnosis is made. After a cancer diagnosis is made, for us to decide what kindof treatment should be offered to these patients. We are living in an era ofpersonalized medicine these days. So all the therapies are targeted therapies.So to understand what is the kind of the tumor profile, that is where themolecular diagnostics comes in handy. So we give...tests which are there in theNCCN guidelines. So based on that, we have these diagnostic tests and the kindof test kits that we use have to be approved for diagnostics. have to be US FDAapproved in certain cases. So these are the, you know, and there are standardSOPs that one needs to follow. And then also the accreditation of thelaboratory. So we have to use everything which is very standard.

So the lab that is offered - offering thistest should also be NABL accredited. So ours is an NABL accredited laboratory. So there are a lot of quality checks that have to happen when we are offeringmolecular diagnostics. These are the tests that, you know, the standard testswe are offering for different kinds of cancers.

Nagaraja Srivatsan (05:48)

What kind of volumes are you talking aboutas you start to do this?

Vijaylakshmi Ramshankar (05:51)

Yeah, so volumes have been pretty highbecause we are at a high volume center. Let me give you an example. For lungcancer, we look for mutations in the EGFR gene. Like if a patient is having amutation in this gene, then they respond to this drug called Gefitinib. Sotherefore, before this drug is administered to the patient, it is important totest for this mutation. So this mutation is being tested in the laboratory.

Earlier we were doing with the qPCR based,which is the same test that was being offered for the COVID. Everybody is awareof qPCR. It was exactly the same test we were offering because that was onlyaffordable for our patients because ours is a charitable hospital. But now wehave evolved into giving next generation sequencing where we are doingmultiplexed sequencing of different panels of genes and based on the mutationsthat are occurring in these genes from the patient's tumor sample, thetreatment is tailored. So this is the kind of evolution that has happened.

Nagaraja Srivatsan (06:50)

You know, as you are doing next generationsequencing and all of that, you must be collecting so much data. Yes. And nowyou're becoming a very data rich in my opinion. Yes. How do you go about managingall of that?

Vijaylakshmi Ramshankar (07:02)

It's going to be very, very challengingbecause now it's not the simple data sets that we were used to in during my PhDdays. Now we are all into the era of big data. So when you have a lot of datato be analyzed, then we need high computing power to do that. And that is whereI think the role of artificial intelligence is helping us in a big way.

So there is a ICMR funded project that I'mworking on in leukemias where we are using AI to develop a model which canpredict relapse in patients. So this is something we are doing it in twophases. So in the phase one, we just take data from the health records, whichis there along with the lifestyle parameters that is also captured in therecord. So all that is taken.

And then we apply the AI based algorithms to predict a recurrence, whether thepatient will undergo relapse, if so, we also generate a score. So this issomething we are doing based on the EHR. In the phase two, we will also add thegenetic factors that we will derive. For example, what fusions are there inthese patients, what mutations are there in these patients, and how that isgoing to impact prognosis. So that information will also get into the AI model tohelp us predict relapse in an accurate way. So this is going to be developed inthe form of an app. So you'll have an app where a clinician, once he enters allthe data, patient-related data, onto the app, then it gives a recurrence code.Then you know that the patient's outcome may be poor. That case, the treatmentcan also be tailored differently. So this is how we are using AI. Just onesmall example of how AI is being used.

Nagaraja Srivatsan (08:42)

Wonderful, because most people say thatwhen you do AI separately to make decisions and then don't bring it within theworkflow, then AI is not well utilized. I think you've bridged the gap betweendiscovery of the recurrence score and then to bring that in the clinicalworkflow to make it happen. Yes. Why don't you talk us through that journey? Itmust not have been easy to bring in EMR or building this AI model. Tell me, didyou do it all in-house? Did you have to do other expertise you had to get fromother people? Walk us through the journey. What was the state before where youhad only EMR records and no AI? How do you go about starting it? And then howdid you get the clinicians to adopt it because that's going to be quite achallenge.

Vijaylakshmi Ramshankar (09:26)

I think I was kind of a bridge between theclinicians and the AI experts. So this is something that we have to do withvery high level of expertise. Luckily, we have the IIT Madras, know, helpingus. The Robert Bosch AI Center of IIT Madras, we're collaborating with them. Sothey have helped us to, you know, write the algorithms, improve on the MLmodels. So the project has, you know, a data engineer, ML engineer who's a partof this project, who's the workhorse who understands. So they're not from themedical background. So my job is to educate them, you know, teach them what thewhole thing is about. And then when it comes to adapting the clinician, themore user friendly it is, the more it is going to aid in their daily practice.Clinicians these days are very open to, you know, AI, using AI in their dailypractice. So therefore, it is just that the best of the minds come together andmy job has been just to bridge these. You know, clinicians are best minds andthen you have the AI, the best mind. My job is as a principal investigator. I'mjust integrating everything together and then trying to see whether it can makean impact in the real world.

Nagaraja Srivatsan (10:35)

Wonderful. And is the AI being used as aprimary endpoint collection tool or a diagnostic tool so that you have to get validation?

Vijaylakshmi Ramshankar (10:43)

This is a kind of an application that wehave made and the models are getting trained with more and more data. The datais from the EMR and then in the phase one we have trained it with there is atrial which is going on in this country which is funded by college - ICCL trialwhich is funded by the ICMR. So we took the pre-trial data of this particulartrial which is integrating the treatment, uniform treatment across the country forleukemia for pediatric cancers, pediatric acute lymphoblastic leukemia. So thetreatment is going to be the same because when you are going to be predictingthe recurrence or relapse, the treatment has to be uniform. So this trial makessure that the treatment is uniform. And then the pre-trial data was taken intothe consideration for developing the model. And now we are working on the trialdata, which has got more number of data sets to kind of sharpen this model moreand more. And then we will add the genetic component also into it. And then wewill make it a complete solution where across the country, whoever is treatingpediatric ALL can use this application to derive a relapse prediction.

Nagaraja Srivatsan (11:50)

Wonderful. And you said standardization ofthe platform is very important. Is there a standard EMR across India or it's very...

Vijaylakshmi Ramshankar (11:57)

There is, there is because the ICMR hasgiven standard proforma. There are standard formats available and we kind of catalog the patient's data as per that particular format. So there we capture everything. So that proforma is used to translate. It's got translated into the application.

Nagaraja Srivatsan (12:15)

Fantastic. And now that in your phase two,you're going to be bringing in more genetic related coding. So that's lab data.

Vijaylakshmi Ramshankar (12:23)

The lab data will, the diagnosticinformation will also get added. The phase one model has got very preliminaryinformation, but in the phase two, we will add very intricate informationpertaining to what fusions, what mutations that we are deriving out of thepatient. And we are going to see, and these are all standard studies which haveshown that if these mutations are there, it may be a standard risk or it maycome under the high risk. These things are already there. So we are just tryingto derive what is known and then we are integrating that into the model so thatthe model is able to predict the relapse.

Nagaraja Srivatsan (12:57)

It's fantastic because as you could see outin the marketplace, people are using more genomics, proteomics and other omicsrelated datasets. Is that your vision to create a multiple omicsinfrastructure?

Vijaylakshmi Ramshankar (13:11)

Yes, think that is a way to go forward andnow after this Genome India project which was launched in January 2020, thisobjective here, the objective was to create a reference genome for India andthat's what has happened. Now slowly this is going to get into cardiovascular,mental illness so we are going to have genetic information for differentdiseases derived; similarly we are also deriving the same thing for cancer. Sohere there are several databases which are being, you know, which are comingup. There is this Indian Cancer Genomic Atlas where people have sequenced thebreast cancers which are there in our country and that data has got deposited.So we want to do this for every cancer which is there so that we will haveIndia specific, you know, databases. When we are into research, we look at thecancer genomic atlas, TCGA databases to understand if we are deriving a gene,we want to understand the implication and outcome, survival. Immediately werefer to the TCGA data set. But those have been derived from the Caucasianpopulation. We need something which is India specific. So all the researchersof the country who working in cancer now are getting together to work towardsthis collective effort of having, you know, deriving cancer specific databasefor Indian researchers.

Nagaraja Srivatsan (14:34)

And is this database being shared acrossthe globe or not globe but across India?

Vijaylakshmi Ramshankar (14:41)

Yes, that is the whole idea becausewherever the data is there, then if you are a researcher, you have to have acertain credential to use the data and there is a committee which decides onthe data access. So, if the researcher or the clinician who wants to make useof this data can justify the reasons why they want to have that access, thenthe data access committee gives them that access and they can access allinformation from the portal which will help them to understand everythingcomprehensively.

Nagaraja Srivatsan (15:12)

Tell me a little bit about the state ofpatient privacy and patient consent in this marketplace. How is that evolving?Because that's a very critical part of having access to the data, which then canbe used by AI to learn. And then, of course, you can then use that for...

Vijaylakshmi Ramshankar (15:27)

Yeah, this is very important and verypertinent question. As and when we have more number of data then the datasecurity is a major concern then it's going to be sensitive patient informationso therefore a lot of ethical know consents have to be in place, ethicalcommittees, approvals have to be in place but majority of the data that goesinto the databases are all encrypted. So you have different ways of encryptingthis data set. So it is not the patient's identity is not revealed. It is onlythe relevant genetic information which gets catalogued into the database. Thosechecks are in place.

Nagaraja Srivatsan (16:05)

Wonderful. And you've talked about the bigdata nature of it. Just how many patient records would be around in this dataset? How much of volume or size are we dealing with?

Vijaylakshmi Ramshankar (16:14)

I think so as of now only for breast theyhave been able to do about 50 to 100 patient data sets has gone into thisdatabase but in the Genome India project they have been able to have aconsiderable number of data sets there. That's the whole idea because we wantto bring together all the researchers who are working on, you know differentcancers. For example, you may have somebody working on lung cancer from somedata memorial hospital, somebody working on lung cancer from Ames. So all theselung cancer researchers are going to come together and going to help with thisinitiative. Whatever sequencing they are going to be doing, it will be auniform protocol that we will use for sequencing across all the centers.Whatever data we derive, and that will get into this database. So it's going tobe a collective effort. So this way, as of now, we don't think we have builtbig numbers in these databases, but we are in the process of doing so.

Nagaraja Srivatsan (17:10)

Wonderful. You said you were the bridgebetween the AI folks and the clinical folks. Walk me through what kind ofchallenges did you have to make the translation? Did the machine learningperson need much more understanding of the domain? What kind of issues did theyface and how did you bridge it?

Vijaylakshmi Ramshankar (17:27)

I think the machine learning personlearning the medical aspects was very challenging because they have absolutelyless knowledge about the actual clinical scenario. The actual clinicalscenario, one can understand only when they work in a clinical environment. Sotherefore, that was challenging. But thanks to the lot of information which isavailable and tools like Chat GPT, which is able to give you very pertinentinformation, that has helped them in building their knowledge base pertainingto the medical issues. So once they are able to understand the problem in alogical way, then they are also able to find solutions in that manner. So thatwas challenging, but we were able to overcome that challenge. From the clinical perspective, from the clinician's point of view, unless something isevidence-based, then it's very hard to convince a clinician. Everything has tobe evidence-based.

And so therefore, we had to createevidences, actually had to, you know, the model had to work in such a way thatit is predicting things that they already know with confidence. Then they startbelieving things which the AI predicts, which the clinicians do not know. Sofirst you have to give them information that they already know, which is comingup beautifully through the model. Then they start trusting the system. So we lookedat the features which are known to be giving bad prognosis, those were comingup in the model. So that gave them the confidence to work on this more. Andtheir inputs have been very, very critical in sharpening this model becausethey exactly know what the mind will expect. They are going to be the endusers. So a clinician there has helped us to decide and design the whole thingin a way that it is acceptable to every other - because they need objectiveinformation. That is provided.

Nagaraja Srivatsan (19:17)

And so you're giving them reinforcedlearning and giving feedback back to the model. It's the human...

Vijaylakshmi Ramshankar (19:23)

I got him together, yes.

Nagaraja Srivatsan (19:25)

And as you start to build these large datasets, the model is going to be evolving. Is that part of the design?

Vijaylakshmi Ramshankar (19:31)

Yes, I think the more and more data setswhich are going to be given into the model, it's going to be robust. So in thisleukemia project, we are trying to do that. But this is not the only thing.There are several other areas where AI will be extremely helpful. For example,in imaging, in areas where we are, let me give you one more example. In amolecular prevention initiative, we do this human papilloma virus screening inthe asymptomatic community women.

So, they are all asymptomatic normal peopleand we are going to derive data whether they are going to be infected with thispapilloma virus there or not. So, the positivity of the virus is about 6.3percent in Tamil Nadu population. Now, these 6.3 women need to be monitored;that is  they are the ones where you knowthe virus is there, but not that every woman who is harboring the virus willinvariably develop cervical cancer, but we need to understand whether they arehigh risk HPV infected, if so, is their, you know, cytologically, what isabnormality they have. So from the same sample, we do a pap staining, and thenit is kind of examined by a pathologist. Now, when we are doing it in acommunity with so many samples, it is humanly not possible for one pathologistto screen so many slides. So this is where we are again exploring AI to, youknow, look at the slides, take the images, and then you training the algorithmto pick up the abnormality. So this is again developed with the help of apathologist.

Nagaraja Srivatsan (21:05)

Fantastic. And is this also you're usingthe AI.

Vijaylakshmi Ramshankar (21:08)

AI, we are, yes, no, here this is we areworking with another commercial company. We are trying to do that. And there'sjust another example I'm giving you where AI is helping us in our day-to-dayoperation. Another area is there is another ICMR funded project where we areworking on surface enhanced Raman spectroscopy, developing the spectralpatterns from the biological samples. So, the patterns that we get between anormal, a pre-cancer and a cancer will have very subtle differences, which isnot very visible to the naked eye. So, we again use AI there to kind ofdemarcate which is normal, which is abnormal. So, that demarcation has workedvery well for us. This kind of spectral patterns can be derived from exfoliatedcells, from blood samples, from saliva samples, any body fluid.

Each of these will have a particularpattern based upon the biochemical composition of that particular biologicalsample. So, we need to identify the subtle differences and that AI is very goodat doing it because we are training those models. So, it will exactly tell youwith the spectral pattern whether it is falling under a normal or whether it isabnormal.

Nagaraja Srivatsan (22:19)

These have been very great use cases aroundusing AI for diagnostics, now AI for imaging, AI for spectral analysis. Wheredo you see AI going in the future?

Vijaylakshmi Ramshankar (22:31)

I think AI is going to have a lot ofapplications in the field of oncology right from multi-omics, where we talkedabout integrating genomics, proteomics, metabolomics. Everything is going toget integrated, and you're going to have a lot of data set. So from this multi-omicsdata sets, if you have to derive the right kind of biomarkers, then AI will bedefinitely useful in this area. The second area is imaging.

Imaging is widely applied these days in thefield of mammogram. So when you have mammography being done, you cannot have aradiologist sitting and seeing every single mammography image. So their AI isbeing widely used. Similarly, it has got application in lung cancer screening.It's a known application in lung cancer screening. So therefore, every low-dosespiral CT scan image can be - AI can be used to pick up early lesions beforethey are becoming cancerous. it can be picked up at early stage becauseanything which is detected early is curable when it pertains to cancer. And anyday, we all know, prevention is better than cure. And when the cancer isdetected early, you also have the treatment is by a single modality. Sotherefore, it also saves cost. So that way, AI is helping in a big way to detectearly - diseases early and including cancer. Then AI can also help in remotemonitoring, for example, with the help of wearables these days. So you havewearables too, and then remotely using telemedicine wearables, it can monitorthe health status of an individual. So that way it is helping in maintaining abetter lifestyle, better acting like a health coach. So these are all veryuseful when it comes to prevention of diseases including cancer.

Nagaraja Srivatsan (24:24)

Yeah, fantastic. I think you've given agood spectrum of AI use cases. As you are in this role in cancer, what do yousee in the next 12 to 24 months? What do you see going to happen in your joband how is it going to impact? How is AI going to impact that?

Vijaylakshmi Ramshankar (24:40)

I think every research project that we areundertaking will have an AI component. The way it is going, it's becoming likethat. For example, if we are doing, say, a hereditary cancer project, we wantto understand the genetic predisposition of cancer in the Indian community. Sowe are going to be looking at, you know, familial incidents. If you are havingcancer in a family, it's running in a family, so we would be, you know,sequencing the patient sample and their immediate relatives and we are going tobe looking at gene mutations and all that. So when it comes to doing this kindof hereditary testing, you want to, not everybody will have access to this kindof hereditary testing. So if someone from a remote place also wants to havethis testing, how do we spread this information? How do we spread awareness?That can happen through an application. Similarly, when someone is narratingtheir family history, then you could have an application which can draw thepedigree chart then and there. So you exactly know and based on what mutationsyou are deriving, it can connect with the different databases which are alreadythere to see whether this variant has given rise to a cancer in a particularcommunity before. If so, what percentage? So all this information can becollated in one place. AI is able to help in that collating information andgiving it in a way that it is precise and concise. So, there will not be anyapplication which will be without AI in future the way it is going.

Nagaraja Srivatsan (26:08)

That is fantastic. Given with India andgiven how cost is a key consideration, how do you see this whole thing aboutdata, data collection, AI models, the cost of it? Can you tell, is itreasonable? there innovative ways in which to reduce that? How do youdemocratize it for a nation like India?

Vijaylakshmi Ramshankar (26:26)

I think India is very good when it comes toits strengths and IT sector. So I think most of the applications that we aredoing, we have the right kind of expertise to do it. And with such a lot ofhuman capital, I don't think it's an expensive proposition for a country likeours because we have the best minds and best kind of skill sets to do thesekind of activities. So I think we are in the best era where when it comes todigital applications, we are evolving.

When it comes to molecular and deriving geneticinformation, we are evolving. And people like us are trying to collate things.And everything that we are doing is interdisciplinary. It is not pertaining tojust one field. We are collaborating and working. So like in my team, I have anML engineer. I have an AI expert, a clinician, a molecular biologist. So we areall working together. So I think this is the best era. And I don't think wheneverybody's going to work together in this way, cost will be a major hindrance.We can still do it in a cost effective way. Indian minds are tuned to the kindof problems we face. I think those problems are helping us to find uniquesolutions as well, which we call as the Indian jugaad. We are very good atthat. So I think that way our problems have helped us to come up withinnovative solutions. And that way I think we will be able to do it in a costeffective way because cost is, as you said, is a concern.

Nagaraja Srivatsan (27:47)

It's fantastic given India's about 18 % ofthe world's population and you're collecting all these samples. Do you thinkthey would have progress towards diversity, making sure that we have a goodrepresentation? We are a wide rich country with wide differences. Are we ableto bring all of that into the data set to me?

Vijaylakshmi Ramshankar (28:05)

Yes, I think we should be able to. incourse of time, we have just started; in course of time, we will have all thisdiversity captured as data. And that data is going to be helpful not only forIndia, but for the whole world pertaining to cancer. If they want to have thisdiverse information, I think India will be in a place to give that diverseinformation to the clinicians and to the researchers across the world. And thatwill help us to come up with innovative treatments, innovative drugs. So wewill be there to provide that information.

Nagaraja Srivatsan (28:35)

This has been really inspiring. I want youto leave us with some key thoughts and takeaway because you've been in theforefront of bringing AI to a very wonderful area like cancer and being able tomake an impact both on prevention and diagnostics. What would be the keytakeaways when somebody is embarking on this journey?

Vijaylakshmi Ramshankar (28:57)

When it comes to AI, you're going to have awealth of information. So we have to always understand that it is augmentedintelligence and not artificial intelligence. So therefore, the human mind todecipher and use the right information for the right problem has to be there.We cannot be blindly accepting the AI models because the AI model functionsbased on what data we are feeding into that. So we have to be sure of what weare feeding in to understand what we are getting. So therefore, the humanintelligence is definitely the one which is driving AI. So that should not beforgotten. We should not be completely driven by the AI base. The human mind,especially when we are dealing with human samples, the human intelligence willbe more important, will have more primary role in getting and deciphering thisinformation to derive useful information which can be translated to the patient.

Nagaraja Srivatsan (29:52)

Thank you again. This was a wonderful conversation. Really appreciate you taking the time to come here. And yeah, waswonderful to see the practical use cases of AI. And definitely, India and whatyou're doing is in some of the forefront of applying AI. So thank you.

Vijaylakshmi Ramshankar (30:08)

Thanks so much. Thank you so much. Thanks for having me and great, wonderful chatting with you.

Daniel Levine (30:13)

Sri, it was a terrific conversation. Whatdid you think?

Nagaraja Srivatsan (30:15)

That was amazing. What I really liked aboutthe conversation was it was very use- case driven. She had gone and talked tous about molecular diagnostics and the apply of AI towards that, which wasfantastic, but also be expanded on the use cases on imaging as well as spectralanalysis. So it was really a good testament on the state of art of what you cando in cancer research and how you can apply AI towards that.

Daniel Levine (30:41)

One of the things you talked about wasusing molecular profiling after a diagnosis to match the right drug to theright patient. What role can AI play in that?

Nagaraja Srivatsan (30:52)

As she had conveyed, I think AI can play avery big role in patient matching. As you can imagine, with large data sets,can create proactive trends and patterns. We can then use that pattern forreally diagnosing care much better. As she said, if you could early diagnose alot of these cancer treatments, then you can actually help solve them from acost-effective standpoint, but also help patient burden or reduce patient burnquite a bit.

Daniel Levine (31:19)

You also asked about all of the datathey're generating and how they're leveraging that. One way they're doing thatis to review electronic health records to predict relapses. These arerelatively small data sets now, but likely to scale very quickly. How big achallenge is not only managing all of that data, but getting the most out of it?

Nagaraja Srivatsan (31:40)

Yeah, I think it's a journey as she said,we're in the beginning part of the journey, but what I really liked about theconversation was a strong collaboration infrastructure which exists. Andtherefore I think the data is going to multiply through the network effect ofthese collaborations. That's one. The number two, she talked about not just EMRdata, but bringing in diagnostic data, lab data and others. And that's gonnareally help scale the infrastructure in terms of both diversity of data butalso the data volumes. And that's both going to be really good to then buildbetter and predictable AI models.

Daniel Levine (32:17)

You also asked about adoption. Are theredifferent challenges in getting doctors to adopt this kind of new technology inIndia than the United States? Do you think they're the same challenges?

Nagaraja Srivatsan (32:29)

I think doctors always want to make surethat they're sure of what use of technology, how they use technology. So it'salways going to be very similar. You have to win their trust and confidence.And you have to win that by providing them effective and trustworthy data or trustworthy answers. And so I think that challenge is going to be there, butmore you start to provide them trustworthy answers, the better it is going tobe from an adoption standpoint.

Daniel Levine (32:56)

Well, it was a great conversation. I'm gladI was able to tune in. until next time.

Nagaraja Srivatsan (33:03)

Thank you, Danny. Take care.

Daniel Levine (33:08)

Thanks again to our sponsor, AgilisiumLabs. Life Sciences DNA is a bi-monthly podcast produced by the Levine MediaGroup with production support from Fullview Media. Be sure to follow us on yourpreferred podcast platform. Music for this podcast is provided courtesy of theJonah Levine Collective. We'd love to hear from you. Pop us a note at danny atlevinemediagroup.com. Life Sciences DNA, I'm Daniel Levine.

Thanks for joining us.