Go with the Flow II Expert Panel Discussion | Hosted by Ralf Kuriyel, Repligen (10/27/22)

Okay, so I think we will go ahead and get started here, and we'll obviously let anybody else drift in as they need to. But I'll go ahead and get started and introduce everybody. So welcome everyone to the final session of this year's Go with the Flow Expert Webinar Series. Today we have brought back all of our panel experts to have an in-depth discussion regarding two key topics. The first will be Process Analytical Technology, or PAT, insights and challenges, and then the second one will be future perspectives of PAT and manufacturing biologics. Today we are joined by Jay West, principal scientist at BMS, who presented biologics concentration measurement using inline flow VPX instruments during the UF/DF process in drug substance manufacturing, sharing the progress that his team has had in implementing the inline flow VPX into their downstream UF/DF for drug substance. Next, we also have Renata Pachinko, associate scientist, and Marius Mueller, lead team lead for analytical fill/finish support from Janssen Pharmaceuticals. Their presentation was on automation of biologics drug product manufacturing through modeling and flow VPE. There they shared how Janssen is forging new ground with model predictive control and how they plan to implement PAT and automation into manufacturing using MPC. And just this week, a couple days ago, we had Ashwaria Ramadan, senior research scientist from Sanofi, who presented implementing PAT strategy using CLPX system for inline monitoring during batch and continuous chromatography. Ashwaria presented her perspective on all of the potential opportunities for the flow VPX system in continuous manufacturing process and work being done in various stages of the purification process. Our expert moderator for today is Ralf Kuriyel, Senior Vice President of Research and Development here at Repligen. Ralph joined Repligen in 2016 as the Senior VP of Research and Development, where he oversees the company's R&D efforts focusing on innovative solutions to address biologic development and manufacturing needs. The structure for today's panel discussion is simple. We'll break it into a couple sections: PAT insights and challenges, and then future perspectives for PAT. For each section, we will ask the expert panel predefined questions to get the conversation going. Throughout this discussion, at any time, please ask your questions using the control panel on the right side of your screen in the GoToWebinar platform. Click on the Question tab and there you can enter your question. At the end of each of these sections, we'll poll the audience for questions and ask all the questions that we've received. Remember to feel free to write in your question at any point in time, and we will be sure to get to your question before going on to the next steps. So without further ado, I want to get this discussion started. I'm going to pass it on to Ralph to kick it off with our Go with the Flow expert panel discussion. Thank you, Ralph. Floor is yours.

Sure. So the problem that we were trying to solve, going back about five years ago during drug product or drug substance development at Bristol-Myers Squibb, was that our biologics, particularly the monoclonal antibodies, the drug substance concentration target was going from 50 to 150 grams per liter. That's a massive increase in drug substance concentration, and the reason we were doing that is for subcutaneous injection, which is a much simpler way of administering these medicines. So it was a quantum leap in terms of the clinic and delivery of these. However, it presented a significant manufacturing challenge, especially in terms of when you're reducing the volume of your drug substance during this operation, and you're relying on weight and the change in weight during the operation. It can be quite inaccurate, and you can overshoot or undershoot. So there was no real way to reliably control the process without monitoring the concentration live inline. And so what happened is we purchased some of these flow VP instruments and started using them in our process development in pilot scale labs. So that was the problem: we had no control or idea of what the accurate concentration was during this process.

I think I'm very much in line with Jay and what they saw. So throughout my career with downstream, I've seen instances where we've had a challenge in optimizing our TFF step. So I think I've always wondered, it would be really nice to have an inline concentration measurement system so we can complete our TFF with better accuracy. And that's where it started. Then when I started doing more and more chromatography, I realized, say first step like a DBC, we're looking at a large number of samples that need to be read offline and also having to wait for the results. The wait time in general and also the amount of sample that you're using up to determine your DBC, I felt it would be really nice to have an inline control or an inline real-time measurement for steps like those. Also, for the most part, the UV detectors max out with any chromatography system, and I think those are times where I felt a variable path length technology would come in very handy.

I can start quickly. So for us, it was similar also because we were also at some quite blind spots in the process. And I think that's basically what we started off in the whole PAT initiative that we were looking into, to really bring also the fill/finish and the drug product process to have online technologies that would enable that. And while mapping this drug product field, we were also looking out where can we use or where can we replace several CQAs that we are currently taking samples for, if we go into and analyze them in a lab in a QC environment. Where could we potentially replace those ones? And eventually we were going out and saying the most critical one, or one of the most important ones, is definitely protein concentration. And we were actually going out and looking at what kind of technologies are there and found things like Terahertz IR, Raman, and so forth, but eventually ended up with the flow VP back then. And found this to be a good, actually quite promising technology that we then basically went forward with, especially also that the technology was the same, also using A280 compared to current offline technologies which use the same, which made it easier also to adopt and to go forward with this one. And with this one, it really helped to basically increase shift, helped to increase sample count that you would have. So basically came off all the benefits that these online technologies have.

And just to add to what Marius just said, the dilution process for which we built our control feedback algorithm is one of the most important steps in drug product manufacturing. In fact, it counts for up to one-third of the product's manufacturing, particularly at the Janssen site. And with that, having the possibility to measure protein concentration, it came naturally to think about a way to automate this dilution process. With a concentration measurement, it's easy to think about.

This actually was, as I said before, also in line with the offline technologies that we were using. So it was really like, I mean, current technologies we have the Solo as well as A280 as comparative technologies. So therefore we basically applied similar criteria and specifications that we also had for those, and then really went forward. As you mentioned, we really looked on the one hand definitely on specification, like on specificity that was basically working, also then range for a certain product, including intermediate precision. But again, for us, what we had seen especially having this new technology was most important to have an offline comparability, also showing a head-to-head with the current standard, which is usually I think not part of classical method validations. But due to this being asked quite a lot, we were basically also having this data included in all the validation work with it.

Yeah, so for me, again like Marius was saying, I think comparability was what I was looking for for starters. So we had the VPX measurements compared with the Solo, which is our typical offline measurement system. And also for something like titer measurements, we compared it with pretty much the platform out there, which is the CXP. So comparability was important for starters. And I think we are looking more for repeatability as well as robustness. Those will be the parameters that I would like to validate.

So in the algorithm we built, there are a few contributions. We have the main part and the core part that is the feedback loop, which basically learns something about the system through the measurements, and based on these measurements and this knowledge of the system, calculates and estimates an optimal action to write back to the system. As I was explaining in my presentation, the complexity of the time delays in the system, for that we evaluated the possibility of having this time delay quantified given the system physical properties and characteristics. But this is not a reliable thing to have for this time delay concept because it can vary due to many factors. As I was mentioning during the presentation, one of these easy to think about is for example the scale of the process, and the scale of the process can greatly impact the time delay. So aiming to quantify the time delay with experiments is something that is really not worth doing because it couldn't provide a sufficiently good estimation of this time delay in new processes. For that reason, we went for a data-driven approach, and it was a different way of thinking to this time delay in our system. So considering the measurements provided by the flow VP and what we expected to see thanks to our mechanistic model, we found a way to make our data-driven algorithms learn while going, while the new measurements were acquired in time.

I want to ask a follow-up question to Renata. So time delay, is that solely a function of a mixing issue with the buffer addition?

The time delay is a function of many things. One of these is for sure the mixing. The mixing efficiency can be impacted by the impeller speed, impeller characteristics, the fluid properties, but also other factors like where is the flow VP placed in our recirculation loop, how long does it take the solution to go to the flow VP, how long does the measurement take. So many, many things that could be quite difficult to quantify and to measure experimentally. So for that reason, we went for the data approach.

Yeah, thanks Ralf. So I put something in the chat, I'm not sure if others can read it, but what something else came into my head that sort of answers your question and the first question you had about the interest in the flow VP, now VPX, in PD is the Donnan effect. If I know people aware of the Donnan effect, like exclusionary, yeah, with excipient drift especially above 150 grams per liter, you get significant excipient drift. Which in our case, like we published, is histidine and sucrose, where when you're getting in the 200 grams per liter range, you can see reductions of 20 to 30% in the concentration of those components. So because you can imagine, it adds another layer of criticality that you want to stop your process, your UF/DF set point, at a consistent concentration. Because if you stop and say at 220 one time or 200, that can make a very significant difference in the concentration of your excipient in that drug substance pool. So that actually added to the importance of being able to push the drug substance to the limit in terms of the highest concentration you can go to without causing precipitation and total loss of CQAs for the protein. And it seems like that limit is around the low 200 range, especially in terms of viscosity, you can't really push these things beyond that point. But by using this inline meter, we can basically safely and consistently have a process for a whole platform of these mAbs at high concentration without risk of the CQAs of the proteins because we're constantly monitoring the concentration. So that's really critical in terms of putting it into the GMP basically, so that we don't have to spend so much time validating the process. Process consistency, we have this thing that's actually going to assure process consistency.

I think this is a spot where adaptive control would come in very handy. And again, reiterating the points that I mentioned for moving forward with a technology like this and moving forward with continuous chromatography, I think most industries are slowly stepping into continuous by moving every step of the process from basically semi-continuous to fully continuous. And I think adaptive control would come in very handy with this transition. Also, I think for starters, from PD scaling up would be the very beginning before it actually goes any higher or any scale up.

It was a multitude of those, right, in the beginning. So first of all, we were checking what problem we could solve. But now that it comes really with, on the one hand, definitely also saving hands-on time, right, which is usually a big driver when you say you can basically have less time that you need for instance an analytical lab over something that might be done on the shop floor. But also we basically have the samples that are not taken anymore from a material perspective, as well as also the soft benefits which would for instance allow to do shifts which would not require to have shifts anymore maybe in a QC environment. But also, and that's also one of the big drivers, is to reduce the hold times for instance for IPC testings, which could be depending also on where and what up to 24 hours, but just the whole time that you would need to include to wait for this result. So I think for at least for the drug product base, the case was quite evident. Pushing into the drug substance space, they were kind of similar, and also the vaccine space, there we had similar drivers where they were saying, yeah, already you know, I mean it was already now when they saw already what you can see now during the process and having this process understanding that they did not have before, already was quite convincing. So this usually helped them to create the case further down.

On this one, it was mostly the question that most of the time came was, how can you have this comparable to what we have? And there we basically used a certain stepwise approach, right, where we first created data from a process monitoring perspective, basically getting a data package there, and then moving it down all the way to really having it as a certain replacement for tests. And the rationale we usually had was, okay, we always showed also offline data, so we always had a certain comparability down the process again. As we discussed also before, as we could refer to the gold standard, the offline data that we also collected during the whole process, with it we basically constructed a really compelling data package. And also we included data we had from our lab scale studies that we did, as well as all the runs that we had at that full scale. So this was really the combined effort which eventually led to approval.

Yeah, so in the pilot lab, it's reducing risk because they have somewhat, I don't know if it's a novel approach, I would say, but along with what Marius was saying, like how do you prove that this is not adding risk to the process, especially in terms of accuracy? And the way we have gotten around that in the pilot lab is that, of course, the flow VP is used or VPX now for process monitoring. However, the Solo VPE is the measurement of record in the batch record. So basically, once you've reached that target concentration with the VP and they manually stop it, it allows to recirculate for quite some time, then pull a sample and then test it offline with the Solo VP. So that's actually the wedge in the door to get this into a GMP space, to say, yeah, exactly, we want to use this for process control monitoring, but then next we want to use it to decide when to stop the process to take a measurement. So that's actually a process like in some of our other PAT instruments that they're used in that way. They basically decide when to stop and pull and take a measurement, which is a baby step, but it's getting there basically to the point where eventually we can trust these instruments to basically do everything on their own in a sense.

I think I don't think we're there yet, thinking all the way there yet.

Yes, I would really say, if you really talk about PAT which would really be to be in control of the process, to really basically have a certain possibility to react on unforeseen things, right, that's basically what we now do. We do see now how the process is, how the process reacts, and this is really something. And I guess Renata can definitely speak on that when it comes to the dilution process, that you can really still interact when you can still stop the process before the process is done and then you take a sample and then eventually you have the risk already. But this comes also I guess with various stages where you see mixing homogeneity, how long does it take in order to really have homogeneity of a certain batch. Getting insights into the whole, I think this really comes with that. And PAT in general plays a critical role to be able to be ahead of rather than just acting but also reacting on things.

Thank you. I totally agree with Marius. It's exactly what we are trying to do with our automated dilution.

Thank you. James, I think sorry, yeah, I just want to add one other point for a really potential use for the VPX that we did not foresee. After the drug substance is done, the formulation is completed, it's put into bags and frozen. And we happened to notice that at these high concentration processes, the concentration of the bags are not consistent and can even vary by 10 to 20% per bag. So because of some sort of settling factor or whatnot, in the future I think it would be quite valuable, and we have some ideas basically to put VPX in line with the bag filling operation to ensure that the concentration is consistent in the frozen bags, because that's an issue.

Oh yes, thank you so much, Ralph, and thank you guys for the discussion. We've had just a couple questions come in here that I just wanted to toss everybody's way before we go on to the next segment in our discussion. The first question that I've had come in is, it seems like everyone is using VPX for various applications. Are all of you using it in the manufacturing stage as well at this point, or is it all in pre-manufacturing? Maybe we go around, we start with Jay because you're on the panel, and then we go to Marius, Ashwaria, and Renata.

Sure. So when they say manufacturing, I think they mean GMP, correct? Yeah, I'm assuming that. So we've been using it at pilot scale for quite some time, which is not GMP but it's a one-quarter clinic scale, and we are in the process of tech transferring it into the GMP space, so not quite there yet.

Okay. And then how about you guys, Marius, Ashwaria, Renata?

Yes, we use it in definitely a manufacturing space, so GMP manufacturing space from clinical to commercial. And right now more and more also the drug substance as well as for vaccines, they are all at manufacturing scale. However, we also have use of it in a lab scale to really accommodate there for studies that we might eventually run on a lab scale. But yeah, so all scales is the simple answer.

Great, thank you. I think we're still in the research and developmental stage, so we're still researching more uses with the VPX.

Okay. And maybe I can say something about the algorithm, the control algorithm we have, and it is not in the manufacturing yet. We are working on bringing it there.

Thank you so much. Another question that did come in was actually asking about method transfers and challenges associated with that. So I think Marius, it sounds like you guys have done it. Did you have any challenges associated with transferring methods with the VPX to the different groups?

Yes, we did. The issue we have is, I think again having the nature of something that is not an offline technology where you basically have an instrument but you can basically say, look, this is the instrument, that's how it's used, that's basically how it's used everywhere, actually gave us quite a challenge because it can be used depending on how the setup will look like for various sites. How it would look for there, we needed to be quite creative there and also seeing how we can leverage things, how we can have things that we might do on a lab scale, how we can have this then on engineering scale or basically in full scale, and also seeing how the equipment differed, how the process would differ to also account for this as well during the transfer.

Sounds good, thank you. I think we got kind of time for one more question before we go over to our next segment here. So the question kind of asks, how are you guys working with your QC groups to align on required CQAs that they do offline? Are all these offline measurements still needed, and are you still pulling samples, or do you also envision ever not needing to pull samples in the future? So I guess we'll go in the same round table section. We'll toss it to Jay, Marius, Renata, then Ashwaria.

Yeah, well specifically in the GMP manufacturing space, as I mentioned, the flow system is not enabled yet. However, it's causing the Solo VPE to be used in an analogous fashion for basically real-time forward processing on the floor. We do that in some of our commercial processes but not yet in our clinical processes, and that's something we want to move forward to. Basically it saves several hours per sample that needs to be tested because the QC method for high concentration samples relies on what's called a gravimetric dilution, it's a very laborious process. And the feedback we've gotten from QC actually is that they're happy to have us take this away from them because it often involves somebody working a third shift because this process typically ends late at night. So someone has to be there late, all these people sign off, it's very stressful, they don't like it. They definitely prefer this something that we take off their hands. So that's the feedback we've gotten from QC.

Great. It's also, as mentioned, during the process we would still keep the samples to really get a head-to-head comparison for as long as it's needed to really get the data package. But once we are there and we have the filing, then we would basically realize that would be then also the benefit to not have QC measuring any samples anymore, to not have these necessities. We still have it as a backup, as a potential backup in case something would basically fall out or would not work during setting this up during any of the manufacturing.

Thank you, Marius. Is there one more question, James, or shall we move to the next section?

Yes, so definitely the flow VPX can play a huge role in moving forward towards these directions because it offers insights into one of the most critical attributes, as Marius was mentioning before, that is the concentration. So whenever we have the availability of this real-time data, we should try to build models, build a deeper understanding of our processes, and try to move into a more objective future where we only rely on what we see in the process at each time point and we take actions based on this real-time knowledge.

Yes, definitely. And I think that's exactly, I mean if you compare the pharmaceutical industry to other industries, right, where you have a lot more of this automation, a lot more of digitalization, I think that bears also kind of really challenges for sure on the one hand, but I think also gives us some chances there right that we can still build them now and build them up and also see how we can have them. Where I think again the PAT really plays a central role in that, to really have also technologies that would enable this, that would be able to react on those, that would by having it combined with any mathematic modeling, with any artificial intelligence, to really react rather on processes than only monitor. And I think this is really where the whole thing is going and where the whole thing basically needs to go. But it needs still also the digital infrastructure, which again I think is in the infancy compared to others, and this is really something to basically align. And as we really talk about PAT, integrated quality, real-time release, I think they always come hand in hand with a proper digitalization, a proper digitalization strategy as well.

Sorry, could you repeat the question again?

Yeah, so I think yes, definitely. The VPX will be a very good instrument in terms of automation, and I think pretty much every step of the process would or may benefit from automation. And I cannot unfortunately speak too much about it at this point because like I said, we're still in developmental stages, but I think it will definitely be a promising way to go.

Sure. Yeah, so my perspective on the VPX is it's a trailblazing technology for enhanced PAT that will serve as a model eventually, I think, for how we, and this is kind of the industry consensus too, because the VPX is based on the Solo VPE, which is based on A280, all very proven technologies. This will be the first of the really enhanced PAT methods that makes a significant impact in the process, and from which all future PAT methods will basically follow that template. And it's a multi-stage evolution.

Yes, I definitely think so. First of all, we should align on what real-time release means. And as Marius mentioned during our discussion, with our automated dilution model, we cannot really finish the dilution and go immediately to the market because there are other steps after the dilution that need to be taken. But the automated dilution can definitely shorten the time of the overall drug product process. And as I was mentioning during my presentation, an estimate of the time currently for doing the dilution process is around 6 to 8 hours, and with our new vision, we could reduce this time to 1 to 2 hours. So it's a big improvement in time terms.

So yeah, I would say that basically, as Renata said before, it's really something where we would need to say what would we say about real-time release, right? And if you say real-time is a driver, I think there it is great to have this. But I think the flow VP is one technology that we might have, but I think there are still more to this that would eventually enable a real-time release. And in the pre-discussion we had, we were also talking about things such as bioburden, which is currently I think one of the tests that takes the longest before having something released. So if you basically would manage to get this time down, or basically if you have the time much more quickly, then other assays are on the spotlight. And I think you can then really just go through that to see which one is then your roadblock to really enable real-time release. And I mean, if you would talk about an ideal case, we would say real-time release should be once the last vial is filled, to be able to have the batch ready and released. So this is I think the ideal case. But I think on that way, if you can already shorten time, if you can really already bring that much earlier, that is definitely helpful on the way there.

I think like Marius and Renata were saying, I think when we're talking about real-time release, we're going to have to define what that exactly means. And like they were saying, I think there are many parameters that go into real-time release, like product quality, bioburden, endotoxin, there's so many characteristics that go in there. And I think every one of these would play a major role when we talk about real-time release. And I think when we have a proven way to show putting all this together, I think will be a challenge going forward.

Thank you, Ashwaria. So I have a, in order to, I've already laid out I think that there's a five-step process or lead, that's what we envisioned to get to full real-time release. And the pressure I see will come from potentially three different areas, particularly in pricing because biosimilars are no doubt already, will become much more common, and that will, you know, the cheaper you can make the drug, obviously real-time release is going to decrease the cost of operations in manufacturing. Another one is the advent of continuous manufacturing. There may be drugs, like already I know in our pipeline, that cannot probably be manufactured reliably without continuous manufacturing just because you get a continuous harvest because certain CQAs will go out of spec if you do a typical batch process. And the third one is contract manufacturing organizations that are manufacturing biologics. They need to tech transfer products very quickly and get the line running and manufacture as quickly as possible to obviously maximize profit. So companies like Lonza are already incorporating a lot of PAT so that they can have basically a platform, hopefully, or eventually maybe real-time release process where they're manufacturing drugs for their customers quickly and efficiently with minimal time for tech transfer.

So it's a relatively simple instrument, and it can be platformized. Precision and accuracy definitely needs to be tested, and you really need to understand the technology and how everything works. Because for example, if you look at the path length, the VPX are getting so short that there's potential, I don't want to slander the product at all, but you can have proteins sticking to the flow cell which can change the accuracy under certain conditions. So it's really something that you need to test thoroughly with all your products to ensure that it's giving you accurate readings under all the conditions that you desire. Because you don't want to, there was actually an issue earlier on with our Solo VPEs because, as we know, they replaced the VP with the X because the need for higher accuracy above 150 grams per liter. So we had our scientists that were not analytical just using the instrument without doing offline testing to see if the comparability was there. So they were just using it without doing the proper comparability testing. So anybody who uses it, definitely make sure you do all that so you don't get poor quality data. Do your homework and make sure everything makes sense.

Yeah, I would say definitely so. For us, it's basically, you know, if I had known what I know now like six years ago, I mean we would have had the process much faster. So it's really also like, you know, there's quite a lot during the process to really implement that. Because I said before, it's not something where you have an analytical tool and you do basically the analytics there, you do the method validation development and then place it there. No, it's much more complex than this. So employing their people, and I think this is really something I would give people along is to have a clear roadmap and also make a proper mapping of stakeholders. Because we have seen for instance that cleanability is a big thing. Then also having the whole engineering and manufacturing there, also have it with people from the shop floor so they know how to operate, to train them, to have the full package around that. I think that is quite a challenge that we might have not foreseen before, but we'll eventually do nowadays.

So yeah, the VPX is definitely a more straightforward system in its mode of operation, and I think the user interface is also very user-friendly. And having said that, I think to understand or to implement the instrument's capability to full use, you're going to have to do a bit more digging into how the instrument works and the full capabilities of the system, and to figure out very exactly in your process where it would fit and what kind of data are the sweet spot for your data generation. I think would need more study into it.

Thank you. And from my side, first of all, my advice is to make sure to learn. I mean, I worked with the flow VP, and my advice is to learn which is the order of the steps to follow in order to not break the cuvettes, because this happened to me a few times. And besides this, to pay attention when someone is working with variable concentrations to these time delays, and therefore to make sure to interpret correctly the measurements when there are variations in the concentration, because it may be that you need to wait a little bit in order to see the right concentration in your process.

Thank you, Ralph. Awesome. Thank you guys so much for the great discussion again. It looks like we have some time for some questions that have come in. And the first question that I see here is, you guys have mentioned that flow VPX is being used at various different points in the bioprocessing sequence. How many systems do you guys envision yourselves using or are using, and can this be instantly moved to different points in the process? I'll toss it to Jay, Marius, and Renata, and not to go around.

Sure. I'd like to have as many as possible, right? Every, well, to be frank, honestly, I think for downstream operations, if you have, in a perfect world, you would want to track the mass throughout the whole operation, right? And that could be done with the VPX, of course. You would get a very accurate mass-to-time plot basically from all the dilutions off your chromatography operations and during UF/DF. But to realistically, I would say, first of all, the low-hanging fruit is the UF/DF. That is where you're going to get the most bang for your buck, especially because the material is completely pure, so we know the accuracy is there, and that really is where it shines and is needed, especially for high concentration operations. So there's a couple, like as I mentioned, actually I'll share a little tip. We've thought this through because even during the UF operation, it would be nice to have it after recovery, during dilution, and then maybe for bag filling. So instead of having three instruments, one way I visualize this could go is that you could, and especially if you don't want to mix your product streams, it's to switch flow cells because they're making available single-use flow cells. So you could have one instrument and just swap out the flow cells as you move along with the process. So you could have the push with the cart to say the polishing operation if you wanted to monitor that for the flow-through, and then move it to the UF/DF just by swapping all these single-use flow cells. So that's a possibility.

Thanks, Jay. How about yourselves, Marius and Renata, any insight there on the question?

Maybe for those who saw Renata's presentation, right, we were also mapping on where would it make sense or where basically is typically concentration measurement for drug product, for finish process. And we also highlighted that there are different kinds of applications where I could think of. It usually also comes a bit on the business case because where we're saying that I think the most benefit that we have at the moment is really using it for the dilution, even then controlling the dilution process. But one could also consider to have it maybe after the filter, right, to also see what the filter impact might be, how much of the filter is absorbed. And then lastly, also directly at the point of fill where it really fills to also monitor the consistency during that. So there are multiple ways where one could apply it. It always just depends I guess also on the business criticality and also on the business case.

Good. And Ashwaria, do you have anything else you wanted to add to this topic?

Yeah, so I think like I shared in my talk a couple of days ago, I think VPX can go pretty much in every process step of the downstream. I think having said that, I think you can definitely toggle the systems around if you were to have multiple systems. And I think the way to go would be probably to identify which step of the process is more complicated and definitely would benefit from an inline measurement system if you want to keep it simple.

Sounds good, sounds good. Thank you so much. So this question is, I guess the person directed it over to Renata, but I think anybody else could potentially chime in and answer as they feel fit and necessary as well. So they wrote, aside from modeling protein concentration, are you looking at other MPC models that you're looking to work with? What kind of would be your vision to be the next bottleneck that you'd want to tackle?

So, well, I am not working anymore with the flow VP now, so probably Marius could answer this question a bit better.

Sure. Maybe on top of it, right, because I mean having this for the control, I mean there are other technologies. And as we talk, all being part of the bio forum, there's a lot also on the Raman technologies in the Raman field there, where it's also a lot on having more chemometrical models that eventually could help throughout the process. And so I think this might also be something where this might apply and definitely help out there. I would say this might be one of the examples.

Sounds good, sounds good. I think one last question here that we have that's come in. The person wrote, bioburden still seems to be a relatively large obstacle that we all face in the process and going through to the... Do you see any new technologies or solutions that will help us get around this test, get around the bioburden process and testing, and get us through some of these?

I know people have been working on it. I'm not intensely involved in it, but there are some clever methods. I know one from years ago that uses the fluorescence, the native fluorescence of like bacteria, which is very common, as an inline bioburden sensor. I'm not sure how far that type of technology has advanced, but it definitely seems that it's something you could validate out, right? Or it's like, obviously they'd like to do the plate reading and all that, but it's so time-consuming that maybe in parallel for several years you could run your inline meters for checking for bioburden and then run your plates, and once it's established that it is a valid method, then we can make that switch. But I think it's possible. And another thing too is like during drug substance manufacturing, there are several sterile filtration steps further down in the operation. So it's basically a very careful process, so the risk of, I'm not sure how often they fail bioburden, but I would doubt it's too often during the process. So should be able to get there, I would hope.

Great, thanks Jay. Did anybody else have anything they wanted to add to that question? I think, Ralph, I wanted to quickly turn it back to you before we wrap up. Did you have anything else you wanted to discuss before we end the seminar?

Thank you so much, Ralph, and thank you to all of our wonderful panelist members here. I think we had a really great discussion for this hour, and I hope all the folks that attended found all of the discussions to be very insightful. So I hope everyone has a great rest of the day, and that will be the conclusion of our webinar. Thank you everyone so much. Thank you, thank you. Bye.