Episode: 7 - Anurag Rathore on Platform Processes, Chromatography, and Expanding Education

Anurag Rathore, professor of chemical engineering at the Indian Institute of Technology Delhi, joins Bioprocessing Unfiltered with host Alois Jungbauer to discuss his long career in advancing bioprocessing engineering. Rathore shares his thoughts on chromatography, using platform processes to accelerate process development, how infinity capture can be a way to arrive at platform processes, and if there is a need for alternative methods. He also explores updates needed to the traditional curriculum of biochemical engineering—and if the next generation should know how to use AI and ML—and his view on bioprocess and development. 

Links from this episode:  

Bioprocessing Summit 
Indian Institute of Technology Delhi 

Bioprocessing Unfiltered: Covering both upstream and downstream processing, analytics, AI and digitization, cell and gene therapy and more, Bioprocessing Unfiltered is your insider’s pass to the researchers tackling—and solving—the day-to-day challenges in the bioprocessing industry. 

Transcript

Welcome And Guest Introduction

Announcement 0:06

Welcome to the Bioprocessing Unfiltered Podcast. Each month we host conversations with the researchers and leaders tackling and solving the day-to-day challenges of the bioprocessing industry.

Where Innovation Still Matters

Alois Jungbauer 0:19

So good afternoon. My name is Alice Junkbauer. I'm a retired professor, but not really retired professor at Boku University in Vienna. And I have the pleasure to welcome you to CHI podcast on Bioprocess Unfiltered. Today we have a very interesting and renowned guest is Dr. Anorat Ratore from Indian Institute of Technology in Delhi, where he is working since 17 years as a as a professor at the Department of Chemical Engineering. Before he was a director at MGIN, adjunct faculty at Washington University, then also a process engineer in pharmacia corporation, and he received his PhD from Yale University with from in 1998. His supervisor was Chaba Horvath at that time, really world-renowned expert in chromatography. Yeah, so therefore, I think this is quite interesting that we have a person who has such a vast experience in the United States, in India, and I think also worldwide. Definitely, we will talk about current needs and skills for training of bioprocess engineering and young faculty, but we let this leave for later. My first question would be where are, because you are a professor in bioprocess engineering, and where are the research needs, shortcomings, but also success stories in bioprocess engineering and biomanufacturing? Please could you elaborate on your own success stories, but also success stories which come to your mind, which have really advanced bioprocess engineering.

PAT And AI For Live Monitoring

Faster Models For Process Control

Anurag Rathore 2:22

Okay. So, first of all, thanks to thanks to Alois for the kind introduction. And so, talking about bioprocessing, so I've been part of the bioprocessing community for now almost close to 30 years. In terms of trends that I see, I think you know, the fundamental work on like say creating a chromatography step or creating even a cell culture step, those, you know, those things probably are, I think, less interesting today from a research perspective because I think already a lot has been done. Of course, there is always new chromatography media, you know, new cell culture media, and those things happen. But from a scientific perspective and scientific innovation, I think the field in general is getting mature from that perspective. Now, where I see as room for innovation and the you know continued continued academic research would be kind of peripheral areas to the core bioprocessing. So, for example, I would say PAT or process analytical technology, which is about you know creating novel approaches to facilitate real-time process monitoring, whether upstream or downstream or midstream. Those, I think that's where there is room for further innovation. Right now, I think there's a lot of excitement in how do we integrate the AIML you know approaches, which is artificial intelligence, machine learning, into our bioprocessing. And the two things I think fit very nicely. We also have been you know working in this area for the last four or five years. And I so I can say that there is a lot of you know, a lot of need for innovation and especially efforts to bring some of these concepts to actual commercial manufacturing, you know. So, because you know, we generate a lot of data and bioprocessing in general. I think there have been more and more sensors, you know, that people are using today, which basically means a huge amount of data is being generated in our facilities. So it makes sense that can I can I you know analyze this data in real time and can I get more and more process and permutation out of this data? And then more importantly, can I create process control schemes which can actually use this information and then take action to control the process in case there is a deviation? So I think that's another key area where I see innovation. Another area where I see is modeling. You know, so modeling obviously has been around for forever, I mean, many, many, many decades. But I think what has changed recently is that because of advances in our computational abilities, we can run these models in a reasonably fast manner. So earlier, let's say 10 years ago, you could not do you know, CFD. You could do CFD, of course. People were doing CFD, but those simulations you know would take you days to finish. And that basically means that I cannot really use it when I'm running a process in the plant, you know. I can't wait a day for seeing my data out of my CFD simulation. Today, I think there are, I mean, people are doing CFD simulations for bioreactors, you know, which which which can be completed in a matter of minutes, you know, so 10 minutes or 20 minutes, you know, that that time range. If if that can be done, now that gives us, you know, a possibility of actually using these models for real-time process control. So I think the same thing can be said about chromatography modeling or ultrafiltration difiltration modeling. So these models now basically the time scale in which these can be run has become small enough for you to actually use this, you know, in a real life setting for making real life. So you're saying also like a metabolic model also could be run. No, no, no, no, so that is one exception. That is one exception.

Alois Jungbauer 6:19

No, just evidence.

Anurag Rathore 6:20

No, no, no, no. So I think there is one exception. That's why, at least so far, whenever you know someone approaches me to get into, you know, like a chosel modeling and things like that, I typically advise them not to go that way because I don't believe, at least as of today, you know, we have models that can be run in this time scale. Plus, to be truthful, I don't think we even have models that can accurately simulate everything. Ultimately, it's just too complex of a system. So, where you get caught is that if you try to simulate the whole cell and every aspect of the cell, the model is so complex that it is probably not even feasible to you know run a complete model, right? So basically it means that you have to take assumptions, right? So most people who are doing this work, you know, they take they make assumptions, and then it becomes a question of that how general that model can be applied for. Because as you make assumptions, you know, there are cases where that model may work wonderfully, but then there will be cases where you know the model will not work better. So that's why that's an exception. I've at least in my opinion, you know, we have not we have not got to that point yet as far as the cell culture modeling is concerned. But like modeling of a bioreactor, modeling of modeling of a bioreactor, as in modeling the hydrodynamics of the bioreactor, the kinetics of the you know, cell culture kinetics, product expression kinetics. I think these are things that can be simulated in near real time today. Not not the cell, you know, cell mechanistic model of the cell per se.

Alois Jungbauer 7:53

You your group really contributed to this model. Give give us a couple of examples, yeah. So it's a okay. A lot of people say modeling, yeah, but this is a very, let's say, vague and broad description. So could you actually elaborate on a couple of examples?

Anurag Rathore 8:11

So I just I'll just give you three, four examples and then then we can go wherever you want. So, second example is where we modeled alpha Laval centrifuge for a mammalian cell culture process. And in that particular case, for example, the cells were, you know, above normal kind of you know vulnerability to shear. So they were having issues when they were running those alpha laval centrifuge at commercial scale, too many cells were getting were lysing. So we did a model to figure out you know that how or what would be the most optimal way to run that centrifuge so that you get you maximize your throughput without you know having too much cell lysis. So that's another practical application of modeling. Third example, I'll take the case of chromatography, where we have created a hybrid model, which is a mechanistic model of chromatography combined with data analytics based modeling, where actually the company is using our model to predict in real time. So based on the charge variant composition in the load material for a cation exchange chromatography column, the model takes that value, you know. So you basically do an HPLC of the load, HPLC fits into the model. The model tells you that next time when you do continuous chromatography, what should be your pooling criteria, right? So it's it's basically allows you to do this real-time control of my chromatography process. And the fourth example I'll give is ultrafiltration dialtration, where basically, based on model of a single you know, SPTFF membrane, we were able to, you know, project at for because when you go to large scale, you have to, you know, normally companies have to use multiple modules. You know, so you have a range of modules available in the market, you know, it could be four and eight and twelve and so on. And the question is what is the most optimal setup so that the total area, membrane area I'm using is minimal, right? So based on that individual a single element analysis, we created basically we coded it's a numerical optimization problem, essentially. And we were able to tell the company that exactly what should be your large-scale setup so that you can get you know enough throughput and minimize your membrane area use. So these are some examples. Yeah.

Who Else Is Leading Modeling

Alois Jungbauer 10:28

Could you tell me from let's say other companies, academic institutions, what what have impressed you the past, let's say, five years or so in this field of modeling and simulation? Yeah. So yeah, I know your your impact is quite substantial. Could you give us also examples, other examples, please?

Anurag Rathore 10:52

Absolutely, absolutely. So I think in general, I would say in all these areas that I talked about, there are obviously many other research groups, you know, that are doing very good work. For example, Professor Sidney in the membrane, you know, side. I think he is also, of course, I mean his his kind of focus is kind of different, right? He's not so focused on the process control side, but in terms of modeling, in terms of optimization, he has done you know some very I know like Bernd Nielsen, you know, he and I, I mean, he's pretty he does a lot of work in chromatography modeling, and I think he very similar work to what we have done. He has also worked with a lot of industry people in that area. And I think, and then there are obviously companies, so Data How, which was came out of you know, Masimo Morbidelli's group. Yeah, they are, I mean, it's a company that is doing exactly what I just talked about. That's their kind of bread and butter. So there are companies coming up now, I think, that are you know offering these kind of services to industry. So I think you know, the many, many, many in people, you know. So there is, I'm forgetting the name, there is a person in Imperial College, Maria. I'm forgetting the last name.

Announcement 12:07

She yeah, yeah, yeah.

Anurag Rathore 12:09

Yeah, she has to have done work in. So I think there are this is kind of an interesting area today for most of us academicians.

Industry 4.0 Versus Digital Twins

Alois Jungbauer 12:16

Yeah, so because because you mentioned Bernard Nielsen, that's actually leads me to continuous biomanufacturing, continuous chromatography. You already mentioned process control because process control is really important when you do a continuous spire manufacturing or continuous manufacturing. Yeah, so there is this industry four dot zero, the digital twins, and so on and so forth. Yeah, could you let's clarify a little bit this nomenclature? Is there a difference between industry 4.0 digital twin? And this is quite often mixed up, continuous manufacturing and so on.

Quick Listener Request

Anurag Rathore 12:59

Yeah, so so industry 4.0 is more of a philosophy and which is which is kind of it's an umbrella which applies to all manufacturing sectors. It's not specific to biopharma, but basically the elements that you know if you if you want to adopt you know industry 4.0, the elements that you need to have in place are basically around you know sensors that can give you real-time information about you know whatever attribute you are measuring or you want to measure. You know, continuous manufacturing is an element because continuous manufacturing obviously gives you higher productivity, and that generally would translate to lower cogs. So that's a piece of it. PAT is a piece of it because it's all about real-time monitoring, real-time control. And then again, adopting this AIML, you know, based, basically advanced data analytics is also part of it. So digital twin is your modeling. That's just you know, kind of a probably a fancier term to for modeling. There's a new age modeling. But the idea is that can I create a model which is so good that essentially it can mimic, you know, my large scale system, basically, right? So if I have a problem at large-scale system, I can use this model to guide me, you know, to kind of come up with a solution to what I should do at my large scale, you know, to solve this problem. So that's a digital twin, basically. So I think everything you said is in a way, I would say that's part of you need all those pieces to actually implement industry 4.2.

Announcement 14:35

Are you enjoying the conversation? We'd love to hear from you. Please subscribe to the podcast and give us a rating. It helps other people find and join the conversation. If you've got speaker or topic ideas, we'd love to hear those too. You can send them in a podcast review.

Continuous Manufacturing For Biosimilars

Alois Jungbauer 14:52

Do you think that this teams 4.2. would also improve biosimilus manufacturing or what would be the content contribution, lowering the costs and and so on? Yeah. So what is your thought about this? Yeah, yeah.

Anurag Rathore 15:11

So yes, I mean the short answer is yes, because like I said, if you adopt continuous biomanufacturing, you can lower cost of manufacturing by you know anywhere from 40 to 60 percent, depending on your process. So, yes, for biosimilar producers, actually it makes a huge difference because you know, for innovator companies, you know, especially companies that are operating out of you know, US and Western Europe, saving on manufacturing cost obviously is important, but to be truthful, not that important. Because if you look at their expenses, the you know, the the amount of the basically the funds that they are spending on manufacturing is actually a very small component of their overall expense. So even if you cut the cogs by 50%, you know, it's not it's not like their profitability will be dramatically improved, you know. So it doesn't really matter that much. I mean, but for a biosimilar producer, it matters a lot because especially biosimilar producers that are based in Asia, it matters because, like, say for a typical biosimilar producer based in India, the cost of manufacturing is almost 60 to 70 percent of the total cost that a company has, right? So if you cut that by half, it obviously will have a dramatic impact on the profitability of that company. So, yes, I mean there is motivation to do this, even though it is difficult to do. And so that's the reason people are here.

Alois Jungbauer 16:35

I think it's a very interesting thought. Yeah, so that you so think that this continuous biomanufacturing and this automation will come first with the biosimulus, then with the originators.

Platform Processes And Product Differences

Anurag Rathore 16:49

Yeah, I so yeah, I don't want to say that because the reason is this that by CVR producing the reason I mean, I just want to give you the reason why why that probably will not happen. The reason is that, you know, so I know of actual companies that I'm working with, right, on exactly this topic. This the problem is this that if you want to implement, say, continuous manufacturing, right? It's not financially feasible to do that in your existing manufacturing plant, which was created for batch manufacturing. Because the the amount of changes that you would need to make will be so much, basically, that it is not it's basically not worth it for you. Plus, there is a challenge because you're using that facility to make commercial products. And so when you make this transition, you would have to stop that facility and kind of you know redo everything. What it basically means is that you have to create new facilities. That is a that is a major hurdle for a biosimular producer because by default, you know, biosimular producers are you know they they are not, they don't have the kind of funding, you know, that say innovator companies have, right? So they are they don't have a tight budget. So for them to create a new facility, even though they know that this is long term, this is profitable, but to actually implement that is not going to be that fast, right? So as of today, obviously the innovators are the ones you know that are further ahead in that sense. But what I was saying was that on the long run, I do think that you know you will see bicimilar producers also adopting that. Yeah.

Alois Jungbauer 18:20

So now maybe let's continue to talk about platform processes. This has been around for for quite a while, yeah. And this is especially for the originators, yeah. They want to have platform processes to accelerate process development and also this actually pressure to be first on the market. Yeah. So what do you think? Have we already arrived for at a stage where we can really talk about platform processes and then take nowadays they say about product agnostic processes. Yeah, what is your thought about that?

Affinity Capture And New Modalities

Anurag Rathore 19:06

Yeah, so at least I mean, so we work, we we have almost like 15-20 companies that we have worked with. And I think by and large, I would say for Moetie, I think the answer depends on the MoIT that you're talking about. Like so for maps, for example, I think by and large, even you know, if you talk to a biosemer producer or you talk to an innovator company across the board, everyone, everyone is following platforms, you know, because you know they don't want to spend a whole lot of time in process development phase, you know, unnecessarily tinkering with the process unless they have to. But having said that, I do think there is always going to be that 10% element where you would need to make some changes to your process because there are some issues that are that that are product specific, you know. So there so there is always, depending on the product, there'll always be a need to make that 10% adjustment. But but I do think that in general companies are following it because not just for process development, but even savings in commercial manufacturing, right? So if you stick to a given format, the amount of like capital, you know, equipment you have to buy or major, you know, new razins that you may have to purchase, stuff like that, is reduced. And so it it really pays off. Now, where this does not work well is obviously when you talk about, say, normal, you know, really competent, like non-MAP proteins, for example, right? But but I think if you talk about a class, like say we have a platform for virus-like particles, and that platform we have seen it works well for, again, virus-like particles. But first time, obviously, you have to create that platform, right? So I think I think the industry in general understands the value on having these platforms, and I think they are investing in creating those platforms because they're worth it on the long run, yeah.

Alois Jungbauer 20:58

Because you mentioned the virus-like particle platform. Do you think that this affinity capture could be a way to arrive at platform processes also for the non-map? Because for AV, the affinity actually is used a lot, yeah. Maybe not so successful because the yield is very low, the costs are very high. What do you think?

Beyond Chromatography In Downstream

Anurag Rathore 21:22

So, I mean, I think for even for because we have spent a lot of time doing this COX modeling. So, from our experience, at least, what I understand is that you know, affinity is fine when you're using it for a product like map, where you are you you're measuring you're manufacturing large amounts of a product and you're gonna have many, many, many cycles of that resin. So then affinity is fine. But you know, affinity is not fine when you have, you know, when you have fewer cycles. And then so for something like AAV, I'm not sure whether affinity, I mean, whether I would recommend somebody using. Affinity for you know that thing because I think that's one example where again, probably the amounts are not as large as maps, right? So, in terms of number of cycles, you know, that you're gonna have may not be large enough to mitigate that extra cost. And also, I think the benefit, like you said, the benefit that you get in the case of monoclonal antibodies, I don't think you get the same level of benefit, you know, when you apply to AAVs. So I think there is room for, and I and one of the problems I see with all these new MoITs, whether VLPs, AAVs, you know, CGTs, all these things is that because these are emerging MOITs, vendors have not really focused, I mean, they have not created solutions specific to these MOITs. So you're kind of applying something that was created for maps or for maybe recombinant proteins, you're applying to this stuff, and then you're finding that they are suboptimal because you know these Mo ITs are physicochemically chemically, you know, very different than maps and and this thing. So I think there is a need for you know for vendors to come up with, say, you know, new new membranes or new chromatography resins which are you know specifically designed for these categories. And I think that that'll probably happen with time. I think in the next four or five years, and it'll happen because maybe the vendors are also looking to see which of these categories are emerging, right? Because you don't want to spend all your time in RD and then maybe tomorrow nobody's making VLPs. Then then you run a loss. So I think it's a wait and see approach, but there is a need. There is a need for novel solutions, you know, for these new models.

Alois Jungbauer 23:40

Because what maybe my next question is there a need for alternative mass also beyond chromatography, filtration, ultrafiltration, like magnetic separations. We have done, I call it alternative, yeah, precipitation, crystallization, and so on. So, what is your thought? Absolutely. Yeah, yeah.

Anurag Rathore 23:59

So, I mean, I have followed your, I mean, we we also replicated your work on precipitation, I think, in my lab, and and I think that works wonderfully. And and I think you know, it so again that's my personal opinion, but for stuff like maps, you know, I felt that it does not save as much, you know, to to for you to deviate from the standard, you know, 2 LC or 3 LC purification. But for these new Moetis, it might work because, like I said, you know, the the the you know you have different volumes that you're talking about, because you know, things like precipitation, you know, or we have done extraction, you know, two-phase separation. So we have done all these things, but I always felt that you know, when you scale them up, you know, beyond a point, it can be challenging and they can underperform compared to chromatography. The beauty of chromatography is you can scale it up to very large scales, and normally it scales very well, which is the reason why it's so popular. But like I said, with these new money OITs, whether it's like mRNA stuff or VLPs, you know, because you're not manufacturing things in those kind of volumes, I think there is value in you know, re-looking at these technologies, technology alternatives. And I think I would not be surprised if we find things, alternatives which are better, you know, than the traditional LC-based apprentice.

Training Bioprocess Engineers For AI

Alois Jungbauer 25:22

So now we have talked quite a lot of new processes, process development, and so on. Now let's come to education and training. Yeah. So we have let's say very traditional curricula for biochemical, chemical engineering. What do you think? Is there a need to change this curricula to update? Of course, we have to update, yeah. Is is the learning by doing outdated, and which skills would we really need, our the next generation would need in context of AI and machine learning and so on. Yeah.

Anurag Rathore 26:05

So we actually at my institute, we just we do a major curriculum review and updating every 10 years. We just finished that. So I'm in the department of chemical engineering, and in our department, you know, we we have made very significant changes to our curriculum. So major things I want to talk about, what we have changed. One is, you know, we there are like four or five courses, I think five courses now that we have going from basics of you know mathematics and numerical optimization to introduction to AI, introduction to ML, AIML applications in chemic. And the idea is that every undergrad that you know graduates from our university, our institute has that at least basic foundation. Now, it it's not, we don't expect, and I don't think that would be the right thing to do, that everybody would become an AIML expert. I personally believe that I think there is still value in you know having the domain expertise. I mean, AIML is fine, but you know, it's a tool basically, it's an approach. You know, not everybody has to be an expert in you know creating new AIML algorithms, right? But I think having, I feel that everybody today, everybody who's graduating at least that age group, you have to be at least AIML aware. You should you should know what it is, you should know how it can be used, right? Even though you are not the coding expert and you know, you're not the one creating new tools, but you should know how to use the existing tools and and what they can do. So I think that's our objective. So we are making sure we do that. Another change we have done is that we have, you know, we are teaching everyone sustainability. So everyone, so sustainability, you know, traditionally in an NOR department at least, you know, we were we we we would not have a lot of focus. I mean, it would be touched in different courses, but you know, that was not the focus. But now we have made it one of the focus. So we make sure that in several of the courses, there is an entire module on sustainability so that again the graduates know the value. I mean, they need to know why it's important, what are the repercussions, you know, things like that. And then finally, another thing we have done is making sure everybody gets some hands-on experience with some of these advanced analytical tools. That's another thing that was kind of missing. So traditionally, we would do, you know, we would have a lab in CRE and transport phenomena, fluid mechanics, heat transfer, mass transfer, all that stuff. But now we are also making sure that at least they know a little bit about spectroscopy, you know, and they have done some hands-on experience. My personal feeling is that for undergrad, you know, we are not looking for expertise. I mean, undergrad, again, it's my personal belief that undergrad should be focused on creating a foundation, right? You you are not creating experts, you're creating a foundation. Masters is where you create expertise, right? Masters is where you can focus on any one aspect. So you, for example, it could be masters in AIML applications and bioprocessing, and and and that's fine. Then you really go deep into AIML, you go deep into bioprocessing, modeling, and you know, you do all that stuff, right? And then, like you, I think another important question you asked was this hands-on training on bioprocessing unit operations. So, at least again, in my experience, it's not feasible to do a lot of that in a normal degree environment. Because you know, the challenge is, and I think in Europe, I guess some of the European universities are fortunate to have you know pilot facilities where students actually get a chance to operate these things. But at least in India, I see that you know it's just too expensive to maintain and create and maintain such a facility purely for training. So it's it becomes a problem, basically. You know, so so rather, what I suggest is that you know, masters is where you kind of focus, you create a foundation in something, but then you can have these shorter, you know, finishing schools kind of thing where you actually, or internships in another good way, where you actually go to industry, and that's where you know you you learn, like say how to operate a you know, fermenter or how to operate a chromatography column. Because creating, you know, a school or you know in a college or institute or university, such a place is very expensive, at least in my experience, I feel it's very expensive to definitely it is definitely yes. We're not able to afford it, I guess. You know, at least here I am not able to justify that kind of expense.

Final Thanks And Closing

Alois Jungbauer 30:38

Yeah, Okay. Yeah, I think we got a really good overview and insight, what you are thinking, what is your view on bioprocessing, research development. Yeah, I'm convinced you are around at different conferences internationally in you in Europe, in India, in the United States. Yeah, and I'm sure those who are interested maybe ask you in person, and I'm definitely sure that you are open also to answer questions if somebody approaches you. Yeah, absolutely. So absolutely I want to thank you for a very interesting discussion and for a very let's say open-minded actually sharing your ideas and vision. So thank you.

Anurag Rathore 31:34

Thank you, and and thank you always for being one of my role models, you know. So I always have looked at you as someone you know who is ahead and you've done tremendous work for the for the for the academia. So I hope I you know I'm trying to copy you.

Alois Jungbauer 31:50

Oh no, no, no, no. I think you go ahead.