Massive Epitope Binning – Infinite Epitope Resolution

Have you ever seen 150,000 sensorgrams?

Watch this webinar on how to analyze larger and larger sets of competitive binding assays, or binning, using the GOLD STANDARD Epitope™ analysis software.

Complete your analysis of tens of thousands of interactions in minutes.

  • Our industry-leading Epitope™ software delivers a wealth of analytical tools to reveal the epitope diversity of large panels of antibodies.
  • Best practices in setting up competitive binning experiments
  • Identify unique epitopes
  • Discover mechanism of action
  • Build your IP portfolio
  • Combine binning and orthogonal data to help prioritize your research resources

This must-see webinar reviews key features in the Epitope software that accelerate therapeutic antibody discovery.

Noah T. Ditto
Technical Product Manager

0:00:00.0 Noah: My main objectives in this webinar are one, to kinda give a quick background on the system/technology that is the Carterra LSA. Particularly we have a lot of new individuals I'm not familiar with on the line, and I wanna make sure everybody's on the same page as far as how the platform works. I'll also move then onto really the question of why bother to carry out epitope characterization, what does it really do in terms of developing new medicines? And lastly, there's sort of latter half of my talk, I'll shift over to a case study of a particular manuscript that had come out earlier this year that had focused on using epitope characterization to understand the Library performance.


LSA is truly a high throughput SPR


0:00:47.3 Noah: So the LSA is truly a high throughput SPR. It was developed with the sole intent to provide a really high throughput straightforward way to screen biotherapeutics, that's the primary driving focus at Carterra. So, without any distractions, that's where we've positioned the instrument, both from a hardware and a software perspective. And at the core of this technology is the... Oops, I'm sorry, I jumped ahead a slide here. Is our flow mobilization routine unit using a 96-channel fluidic head to address the biosensing chip surface. So in this image here, we have a 96-channel head docked on a biosensing chip surface, and it deposits 96 discrete molecules, if you will, onto the surface at a time, and in forwardness to prints we create a three-four array. The instrument then can automatically switch into a single-channel mode where we have a single injection passed across this entire array, is what we termed one against many or one on many.


0:02:00.9 Noah: So with this fluidic setup, which is totally different than any other system that's previously been designed, we're able to get a tremendous amount of data per injection, and additionally, both the 96 and single-channel mode fluidics operate in a bidirectional ejection fashion. So we have a fixed volume of injection we use, flow it bidirectionally across the surface during our association or capturing or what have you, events, and really get the most out of that sample without having to compromise by using smaller slower flow rates or other sort of drawbacks that traditionally have been in the SPR space for testing really low, low concentration materials. We do achieve 384 unique reaction spots that is independent of the 48 reference inner spots. So you do measure 384 unique species in the assay. It's not 384 minus some number of reference locations.


0:03:05.2 Noah: And this really... What this really means is, this instrument operating in a totally different fashion, and all of their systems have means that label free biosensing can be done very early in the stages of drug discovery and development, or previously kind of low resolution plate-based methods typically were the way things were done, because they could screen the hundreds to thousands of potential clones or candidates that were required at those stages. Now really the LSA affords the opportunity to think differently about this workflow and consider the fact that if we wanna do really robust kinetic screening upfront and epitope characterization and other sources of high throughput assays, it's much more amenable now than it had been in the past.


Characterizing Epitopes


0:03:52.6 Noah: And although the focus of my talk today is primarily on the considerations for characterizing epitopes, I do wanna just point out that the two major applications of the system are high throughput kinetics and high throughput epitope binning. So, this is just a quick slide to give all the attendees a flavor of what you can achieve in about a nine-hour run on the system. So 384 unique affinities all measured in parallel against the same antigen injection, so there's no change in antigen properties over time, as we're running this assay. And a very straightforward, very small amount of antigen, so this is a typical output you would get from a single nine-hour kinetic run.


0:04:32.2 Noah: And I'll just plug right now, we have a great webinar coming up next week, details are coming up today, I believe. That's gonna be given by Dan Bedinger, who is our lead application scientist at Carterra, on more details about the kinetic workflows and how to optimize those assays on the LSA. So that should be next week Thursday, so just keep an eye out for an email update on that. But switching back to LSA epitope binning, this slide is meant to just convey a very high level conceptual view of how the LSA operates in terms of epitope binning. So, at a fundamental level, when we're doing competitive epitope binning, we first need to compete the clones against each other, so this is to effectively determine whether or not two clones can form a trimolecular complex in the presence of antigen, and based on these pairwise competitive events that we carry out repeatedly, throughout the panels of 384, we developed these sensorgram profiles shown in the center of your screen here.


0:05:28.7 Noah: So in these sensorgram profiles, this for example, is a classical sandwich-style assay format. We're measuring an increase in response related to antigen binding to a clone on the surface, and then a subsequent either response as the secondary clone is able to sandwich or lack of response, suggesting that the epitopes are shared between these two clones and there's no available epitope to bind. So we can generate these thousands of interactions in a single experiment and cumulatively what we do is, then move on to our proprietary tools we use in the software to represent this. So graphically speaking, we do heat maps, dendrograms, and network plot arrangements that condense down all of these sensorgram signals into something that's very straightforward and easy to understand, and as I'll show in some later steps, we can take these views and even build upon these further with additional attributes at the clones.


 Carterra LSA working on the front lines against the COVID-19


0:06:36.4 Noah: And it's very contemporary to point out that we do have the Carterra LSA working on the front lines against the COVID-19 virus. So, we have multiple labs right now that are using the system to develop both therapeutic and vaccine-related solutions. And I just highlighted one of the, I guess you could call them groups here or institutes, La Jolla Institute, who is a consortium of a number of different entities, including Gates Foundation, and their really mandate here, in the near term is to develop a therapeutic antibody cocktail. And this is a great quote from Dr. Ollmann Saphire at the Institute, really calling out that... While it's interesting to know that they do these antibodies that we're identifying via B cells and all sorts of other sources, bind the target and maybe have a certain affinity. What we really need to understand in order to develop an effective therapeutic cocktail is what is the mechanism of action and how different antibodies complement each other? And what this really speaks to me is that we need to understand where they bind on the antigen and how they relate to each other with the binding, which in a nutshell is what we achieve with epitope characterization.


0:07:54.3 Noah: So why is the LSA such a beneficial tool in epitope characterization? If you're familiar with the space, you can appreciate that there's several different platforms over the years that have claimed the capabilities to do epitope binning and characterization, but I would say that the LSA stands out uniquely for a number of very important reasons, one being the sheer throughput of the system, even the closest sort of label free biosensor out there that can do epitope binning is about an order of magnitude less in terms of throughput than the LSA. Running at peak capacity, you can screen out 384 unique clones on the LSA, and you're going over 147,000 interactions tested in that assay a single time.


0:08:38.4 Noah: In terms of resource requirements, I briefly brought up how the instrument operates in terms of the bidirectional flow that enhances sensitivity, requires less sample, and when each injection passes across the array of 34, that means we get a lot of data back with very little sample. So in terms of a per clone consumption, it's about 10 micrograms at most, to do a standard epitope characterization study. And then, simple set up, the instrument is high throughput, but not via any external robotic configurations or anything along those lines, rather, we just need a plate of clones to immobilize onto our censorship surface, a plate of clones to inject as analytes across the surface to assess competition and a tube of antigen to assess whether the clones can bind simultaneously or not.


0:09:31.0 Noah: And then, sort of the icing on the cake is the epitope analysis software. So again, just like the instrument was designed purposefully to screen biotherapeutics, we've also developed this epitope software with a sole focus of characterizing epitope binning data. So it's built both to quickly take up thousands of interactions at a single pass and make processing very streamlined, as we'll see later in this talk. And also, give you a lot of graphical cues, which I kind of alluded to earlier, that make understanding the data very simple and taken from being something that you need to be highly skilled and understanding SPR signals to a sort of a language that can be translated across many working groups.


0:10:17.0 Noah: And I know we have a number of current customers on this call, so I just wanna give them all a heads up that we do have a new version of epitope that has come out earlier in the year. So I encourage everyone who's a customer right now to go ahead to our website and grab that new build of the epitope software.


OmniChicken Platform


0:10:36.6 Noah: Okay, so on to sort of our case study that I mentioned earlier. So today, I wanted to talk about the OmniChicken platform, and in particular, how the LSA had been used to interrogate this platform. This study was done by Ligand Pharma out in California. And they have... If anybody doesn't know, they have a chicken... A transgenic chicken platform, they term the OmniChicken, that has a number of advantages being non-mammalian, kind of different ability to raise antibodies against difficult targets, for example. But this is a great test case here to understand really what does epitope characterization tell us, and we can kinda see based on this, really where the value is in epitope characterization using the Carterra LSA. So, in this particular manuscript publish... This was published in late January of this year of 2020. There were two main contributions of the Carterra LSA in the manuscript, one was to do epitope analysis via a classical binning essay, which I'll go into a bit today in terms of data. And which is interesting, they also included a domain mapping exercise within that binning experiment. So, including sub-domains of the target antigen in question, which is progranulin in order to both identify groups of antibodies that we're targeting similar epitopes, and then understand where those epitopes resided on the overall protein structure.


0:12:05.7 Noah: And then the second facet, which I won't go into much detail today, is the binding affinity assays run. So this was just a straight capture of the clones in this experiment and running the antigen across the array to assess binding affinities, get really good quality kinetics upfront.


0:12:23.4 Noah: And I encourage everybody to go ahead and read this manuscript, it's really great, very interesting read. Really showcases some pretty cool technology that are developing at Ligand.


0:12:33.9 Noah: So, how was the LSA leveraged by this team, at Ligand to evaluate the OmniChicken platform. So one, the team went ahead and did analysis directly from soups using minimal amounts of clones. So, kind of in the typical workflow where we're generating hundreds of candidates, purification and scale-up resources are often limited and we just wanna know, is this a good clone, is it worth taking forward, what are the properties of this clone? So the team went ahead and did a low kind of matrix expression system to generate each clone, and on the purposes of what the LSA needed to conduct, these three main facets of characterization being epitope mapping, epitope binning and kinetics, there's about 10 microgram in total needed for each clone. So, just at a high level, very reasonable starting conditions in terms of feasibility for any group.


0:13:28.9 Noah: To incorporate the LSA early in the process. And then in terms of the 384-spot capacity, the capacity itself was leveraged by putting a ton of different controls and unknowns and standards onto the surface, so that per each injection, there was an enormous amount of data generated and it could really inform the detail that was going on with these particular clones. So there was about 200 unknown scFv-Fc soups that were arrayed on the surface. There was previously characterized reference antibodies, 20 of those, that were also arrayed, seven antigenic variants. So the antigen itself was coupled to the surface along with various mutants and chimeras. These aided in the epitope mapping exercise.


0:14:22.6 Noah: And then lastly, there were positive and negative controls included. And various molecules can be included as positive and negative control. Some of them can be some things like antihist to confirm that we're injecting the his-tag antigen, for example. Other can be... Positive controls could be anti-FC reagents just to confirm that the scFv-Fc was actually flowing across the surface at a reasonable concentration that could be detected because I believe in this study, there was no titering done or normalization, I should say, of the supernatants prior to introduction into the LSA. They were used directly without any normalization steps to adjust concentration. So we do include those positives just to be sure that we are in fact having a reasonable level of the clone introduced across the array at a time. And then obviously negative controls are great just to check for non-specific background signals. So in addition to our reference surfaces that are already inherently built into the array, you can also include irrelevant antibodies that should have no reactivity towards the target, for example, to really have high confidence. So you can see there's pretty much a kitchen sink on the surface of this array right now in this setup.


 Speed Factor Comes into Play


0:15:37.2 Noah: And then really the speed factor comes into play. So the last bullet here, 220 kinetic affinities measured in a more or less an overnight run, a little less, but we'll effectively just call it you started at the end of the day and the next morning, you have all the genetic affinities wrapped up. And about 48,000 binnings/mapping interactions were done in about four days, give or take. So within a week, you understand quickly what were the outputs of this library. Where do things fall? Did we get the diversity we wanted? Did we get the affinity ranges that we were looking for? So a lot of great information and you don't have to wait a month or multiple months to find out.


0:16:17.4 Noah: So with that, I'm gonna go ahead and jump over to a portion of the data set from these experiments. So this is just the epitope binning, some of the epitope binning for a subset of the clones that were used in these studies. So I just wanna highlight a bit of the software in terms of key features. I know there's certainly people on the call who've never seen the software before, so hopefully you can get a high-level view of how things work. And also, I'll try to point out, as we go along, some of the new updates in this more recent build, for some of those savvier users on the call who wanna check out the new options that we have. Okay, so I'm pulling up our epitope software right now, and hopefully it's refreshing on everybody's screen successfully. So this is sort of when you load in the epitope data into the Carterra epitope software. This is what you get on the initial data page, data tab, which is what we're on right now. We have a list of ligands, which... This is a technically convoluted term, but means species on the surface in a label-free biosensing experiment, analytes, which are injections across this surface in the experiment, and then unique clones listed here, so the total unique species compared in the study.


0:17:35.9 Noah: And on the right hand side here, we have our sensorgrams and this is representing all the cycles during the experiment. So if we zoom in a bit, we can kinda get a flavor for each individual cycle and what was happening. So in each cycle, there is a... This is an antigen injection followed by an antibody injection, and these blue curves here are control injections of just the antigen alone. So this is what we kinda used to qualify the activity of the surface over time. And then our software, there's automated functions that check for surface quality and activity, and if there are drops in activity over time, the software will put sort of gates on data analysis to make sure that we're not analyzing ligand or clones on the surface that are in fact maybe not sufficiently active enough in binding antigen to be useful in the assay. And for the current users out there, one of the nice little adds we got in this latest build was the inclusion of these partial block indicators. So we print in blocks of 40, of 96, excuse me, and we do four events, up to four events to achieve a 384-array, but if we're only using part of the block, instead of just like a check for on and off, we have this partial dash that kind of shows you that you're not using quite all that block, you're using some of it.


0:19:00.0 Noah: And one other thing on this data tab before we really kick off the analysis, I wanna go to the info tab. So I've pre-loaded some data for these clones. We term this, orthogonal data. So while we can... Without any other external data describe these antibodies in the context of their competitive relationships and relate that to epitope, we can also introduce data from other sources, and eventually, you'll see later in the demonstration, decorate the data with other attributes. So in this case, the library source, for example, is one of the attributes. So, simply, we pasted in data here. It's a very straightforward interface of describing the data and then pasting in for all clones. Types of data, you can see standard, for example, and the chain descriptions on here. And then another function is the mapping data. So this was using some of these sub-domains of progranulin and assessing whether or not the antibody is bound to them. So we incorporated that as well. So these A, B, C, D, E are relating to different sub-domains of progranulin. And that's all just added in here, the user has sort of unlimited space to keep adding these and call them up in the data later on, when the analysis progresses further.


Go-Ahead to This Batch Analysis Window


0:20:17.8 Noah: So what we're gonna do is go-ahead to this batch analysis window, in this window here, we go ahead and define basic processing steps. So it's kind of a one and done operation. We're looking at the sensograms here on an overlay type view, so we have a sort of baseline, we have injection of antigen, and admittedly, the scales are kind of squished here, so the data might be a little hard to see on your screens. And then we have an injection where the antibody is and then some regeneration pulses near the end of the cycle to remove any bound antibody and antigen and restart and re-set the surface basically, for the next cycle. So these bars are in position to tell the software, one, where to sort of wireline, crop the data, also perform the initial processing steps to identify the bound levels of antigen, because if you remember, that's an important metric that we track, to make sure that we're binding antigen over the course of the experiment, if we fail to... On a given spot, fail to start binding antigen, we flag that data and by default, take it out, we can add it back if there is...


0:21:25.3 Noah: The user wants to customize the settings. But by default, the software removes it, so there's no confusion about misinterpretation of data. And then we have a second sort of position here that we adjust, relating to the sandwiching antibody response. So at this point in the sensogram, we're telling the software is wanting to know what's the signal of an antibody or a clone binding to the surface in the presence of the antigen. Does it bind or does it not bind, and if it does bind, that might be indication of a unique epitope compared to the clone on the surface, and this is the competitive nature that we iterate pairwise, in order to generate our big picture of the antibody relationships in this panel.


0:22:07.1 Noah: One thing some of the kind of savvy users on the line might recognize is, normally we set this bar towards the end of the antibody injection where we have the maximal signal. I did move it over slightly in this case, to where the injection of the antibody ends and we go back to buffer our sort of dissociation phase. That's because these were supernatant screwed samples, and there is a refractive index difference between the crude sample and the system running buffer, which is HBS in this case.


 Simply Measure the Binding Signal of the Antibody


0:22:38.0 Noah: And in order to overcome that, we just simply measure the binding signal of the antibody when we're in our typical buffer, as opposed to the difference or refractive index of the supernatants. So, that slit over here, to help out the analysis. And with that, we can start the process. So hit start here, it's gonna go through and do some basic data processing steps, and go on and do the actual analysis. So this is curating the data, sorting the heat map, and generating network plots based on what we have here. There is a separate manual function for processing here, as well. So if the user did wanna go in step by step and achieve these, they could do that, but you can see... So for the 384 odd locations on here, of which a subset of them are actually active and being used in the assay, it was on the order of seconds that it generated a pretty good starting point for this data set. So we can see, we have our sensogram on the left here, showing the antigen-binding phase, and then followed by an antibody injection phase, where we have several good sandwichers. As I hover my mouse over, hopefully, it's clear to everybody on the call that we have a clone description on the surface. So this is clone 75 as it's termed here, and then the second value in there is the clone and solution.


0:23:58.1 Noah: So we're comparing those two. And then we also have our blue controls, of our antigen only. So basically, this is antigen mixed with Buffer, as opposed to as opposed to an antibody injection, and this kind of is our starting point. We wanna understand how much above the antigen-only signal are we seeing an antibody response. This yellow line here is what we use to define what we call a threshold, and if it's clear to everybody on this call, there's a green sort of background up here, that's what we consider sandwiching and actually relates to the green sandwiching and our heat map. And a red sort of background down here, which relates to outcomes that we're considering blocking and effectively by saying blocking, we're saying that these two clones have epitopes that appear to overlap in this particular combination. We have quite a large heat map here. If we zoom in a bit on the heat map, we can see that that orthogonal data that I added in previously, is actually being already represented here, and so we're showing the library source and our domain mapping information on top of the heat map headers for each clone. We scroll down a bit.


0:25:14.6 Noah: See, even our standards are in here, so if we click on this, this is one of our control antibodies. That's been previously characterized and we understand what domain it binds to and some other properties. And then it's being compared to some of the unknowns as well. The supernatant scFv-Fc's. So typically, the user would go through, understand the thresholds, the thresholds are adjustable either at a global level or by physically moving them within the sensogram for each species, and to orient everybody as well. The rows in this heat map refer to clones that are on the surface of the sensor chip and the columns represent species that are injected on top of those, on top of the array. So, I'll zoom out a bit on our heat map, and then we can go over to the network plot here. So on the network plot, we have... What we're on right now first, is these bins, and these bins are derived by simply looking at the competitive outcomes across this...


0:26:20.3 Noah: This row in the Sensogram, in determining who's similar to who and who isn't. So in the network plot, either a circle or a square is considered... We call that a node and that's a clone in the assay, and the lines between them represent blocking relationships. So we can see that we have several clusters of clones that sort of group together very tightly because they probably are very similar in who they compete with and who they don't compete with. So our algorithm defines relatedness sort of like how you would relate social networks in terms of similarity of individuals based on who they communicate with or know, in ways like that. And then we have some clones... Actually, this is a great control right here, anti-His, which should not compete with anybody, it's simply an anti-His antibody that's used to verify that we have antigen injections during the array. So a great control though, it should not have any competition.


0:27:16.3 Noah: And if we click on each of these nodes, we can actually see highlighting. It might be a little hard to see on the screen here, but we can see highlighting in our heat map as well, as I click here. So it's highlighting particular clones. If the clone is square, it means we probably didn't have enough data on this as the clone was on the surface, to really determine a confident signal, so it's excluded. We call these uni-directionals, but typically, since we have injection-phase data, we don't concern ourselves too much with that. We obviously want competition in both orientations, but as long as uni-directionality isn't too excessive, it's manageable. And then we can click on... And if I zoom in a bit here, we click on these blocking relationships. It actually brings up the curve, so it'll bring up the individual Sensogram for that specific interaction between these two. So we can kind of compare those and understand if we see two species blocking, what's going on between those species.


Tools That Are Typically Used to Understand Overall Assay Quality and Interesting


0:28:14.9 Noah: And to some of the experienced users on the call, we have some tools that are typically used to understand overall assay quality and interesting, nuanced behavior. So asymmetry is where we get competition that doesn't recapitulate in each orientation, so Clone A is on the surface and Clone B is in solution and they compete. When we reverse that, with Clone B on the surface and Clone A in solution, they should also compete. If they don't, we have the ability to highlight that in the heat map by selecting "Show Asymmetries", and it's a great way to understand, one, that you make sure you're setting your thresholds correctly, but also two, and I'll toggle it so maybe it's visible to everybody on the call. I'm not sure if it'll really show up. It's sort of these lightened boxes here. They become hashed as you zoom in a bit. It's showing species that may be borderline, so maybe thresholds need to be adjusted.


0:29:07.9 Noah: Or possibly, there's some really interesting behaviors where a certain antibody, the order of binding does influence presented epitopes. So a really powerful tool and we threw it right up here on the toolbar, just to make it super obvious. There's also functions that you can graph buffers. So the control buffer injections, the control antigen injections, you can plot those to understand where the signal falls relative to the control. And selecting symmetrical pairs allows you to pick an antibody, an injection, and compare the two of them together, side by side. So I've selected this particular cell here. And if we can find it here, I believe it's right up here, its corresponding alternate orientation is also highlighted simultaneously. So these are great tools to... Once you have this master map to really drill it out and understand some of the detailed behaviors they're seeing.


0:30:01.1 Noah: Switching back to the network plots quickly, we have the community relationship shown in this tab here. And what the communities are, as they are derived from hierarchical clustering analysis of this heat map, so we take the heat map and we generate a dendrogram, which basically builds upon ignoring subtle, small differences between clones and gets us to a point sometimes, where we can find major trends in the data without being lost in individual... At the bin level, because if we just... Conceptually thinking about this, if we binned 200 antibodies and we get back 200 bins, we really didn't achieve anything. The goal is to basically distill down the antibodies, based on epitope into major groups that are highly similar enough, that we can make informed decisions on them. So that's sort of what the communities do. It takes out some of the challenges with partially overlapping epitopes and borderline signals and says if an antibody is largely competing with another antibody and they share high similarity in terms of competition, we cluster them together on this dendrogram, and then we adjust this height bar to define a new way to relate them. So a new coloring scheme. So if we apply this...


0:31:20.4 Noah: We'll just refresh our communities here. And maybe I did pull it up just a bit here, let's see, to maybe get some of these... It's like a lot of them are fairly conserved. We can start to see that it collapses down, these communities, or these bins, into the larger communities, and if I just toggle back, we can kind of see. We had, with example clone 40, and example Clone 9, and they were very similar, competing with each other, competing with basically all the same... And at this level of stringency on the dendrogram, we're effectively able to communicate confidently that yes, they are in fact likely hitting almost the identical epitope.


0:32:00.8 Noah: So the bins give you high granularity and the communities let you get that high view of the data and really understand, for example, this is our major community here, and really understand the big picture of the data as well. So you have that ability to drive down as much as you want to and also step back and make a clear choice on what your major epitope outcomes are. So we previously had added in that orthogonal data to this set, and I wanted to show everybody online, how to pull that out of the experiment. So if we go to node color in our gear menus, and just for everyone's benefit, these gear menus are tremendously valuable. It looks like our software is fairly simplistic, and there's just a few options. If you really wanna get into really advanced analysis, these gear menus have just about anything you can think of in them, from just changing the simple coloring schemes and the graphics, to actually going in and adding different pieces of datas.


0:32:54.0 Noah: So node color, for example. We'll choose library, for example, and now our data, we've maintained the same epitope competitive structure that we previously had, so we're not hiding any of that, but we're adding an additional component that shows us how our standards relate to the different sources of clones from this OmniChicken platform, for example. And you can see that it's great, we have good coverage of our yellow standards across many of the species, and within that, we have diversity of different library sources and in those as well. So this is a really powerful tool, and we've had a number of customers really leverage this, to understand where the antibodies are coming from and whether or not they're getting the diversity they need, to get a really well-balanced panel of antibodies for whatever their therapeutic objective is.


 Going Ahead With Our Mapping Data


0:33:43.5 Noah: I can do the same, going ahead with our mapping data, for example. So this is saying, of the sub-domains that were included here and that we showed binding to which sub-domain are we getting reactivity to for each of these groups. And what's really great about this is, you can see that by and large, there's excellent agreement between the community groupings and the mapping to specific domains, which is expected. We would expect antibodies to compete by binding to the same domain, and therefore should get re-capitulated with the mapping exercise, and sure enough, there's very good agreement. There's only a few bins where you see some interesting or communities, we see a few interesting, sort of carry-over, bleed-over into another bin or community, for example.


0:34:29.9 Noah: It's really, really powerful stuff and these types of images are great. I think... Coming from a drug discovery background, you're often faced with presenting data to a lot of individuals who really don't understand SPR, they really maybe not even fully understand the concepts of the assay you've run and certainly don't understand interpreting the signals. But to show an image like this, is so powerful. You take in all sorts of data and put it into an image that can easily describe what bins we wanna be in and where we wanna move forward, and you can pull a lot of other data into this as well. So we could obviously represent affinity. So within a given bin, if we found that we were interested, say that we had a functionally relevant bin, we could dial in on a specific affinity within that bin. So there's lots of other powerful ways to use these, aside from just representing a direct epitope.


0:35:21.9 Noah: And it looks like we're about running up to the end of the time I had allotted. This was just a quick introduction to our software, and hopefully everybody has a pretty good idea of what type of capabilities the LSA has, for epitope characterization. As I had said, it was really, really powerful in this manuscript that was generated by the Ligand Pharma to look at their OmniChicken platform. Really gave them a lot of great information and did it in a very quick and straightforward way. And with that, I'd like to go ahead and thank everybody for attending and wish everybody the best of health and staying safe, and please don't hesitate to reach out to us. My email's right here on the screen, you can reach out to me directly. If you want some more general information, we have an info@Carterrabio where you can reach out. And again, I'll just take the opportunity to just give my colleague, Dan Bedinger, another plug that he's got a great webinar coming up next week, looking at allocate LSA kinetics, which I hope everybody will be able to attend as well. So with that, thank you. And John, do we have some questions I can answer?


0:36:32.1 John: Hey, Noah, thank you very much. Yep, yep. We've got some questions coming in. And... Here, let me bring up the first one. Do competitive binning assays tell you anything more than which antibodies group together?


0:36:49.4 Noah: Yes. So if you strictly competed antibodies with no other known characteristics in terms of epitope, they would be sort of... You would group them, but you would not have a relationship between them in terms of where do they relate to binding on the antigen structure, for example. Certainly, if you can introduce a known antibody, a benchmark competitor, whatever you wanna call it, that helps 'cause there's typically a soft structure for that. Other ways to do it would be, to include a receptor. So in this particular target, progranulin, I don't believe... The last time I checked, there's a known receptor to it, but if you did have a pathway receptor, for example, you can include that as one of your analytes in the binning exercise, so you get competition data from it. And obviously, if there's a structure showing a solved relationship between the receptor binding site on the antigen or a target, you can infer from that, kind of where the epitopes are laying particularly because you understand if they block that interaction, it's likely that they're doing it in this fashion and localizing close to that interface.


0:37:54.2 Noah: And then if it wasn't obvious in the data set I've shown here, you can use subdomains of the protein, in order to sort of map as you go along within the binning exercise. You can either do that independently or as just shown here, integrate the chimeras or sub-domains, so that you can map them in addition to the binning data. It gets you good confidence of grouping antibodies based on bin but also then taking that bin and actually having a structural location that you can assign to it, which is tremendously helpful in these sort of mechanism of action studies and all sorts of other downstream data that you need.


What types of epitope assays are customers running, to study COVID-19?



0:38:33.6 John: Thanks, Noah. Here's another one. What types of epitope assays are customers running, to study COVID-19?


0:38:46.2 Noah: Yes. So there's quite a few. I think, right now, the big push is to get therapeutic cocktails put together. So in that regard, it's a lot of straightforward binning, to understand kind of what I described here, to kind of understand what groups of antibodies are there, and in conjunction with, say, neutralization assays understand potentially what combinations neutralize the best, and then looking back, having that ability to understand from a structural level, "Okay, these neutralized the best because they hit different epitopes, regions that are vulnerable," for example. So that's probably one of the big pushes right now, but there's been some great work previously, generated by groups at UPenn, the Cohen lab, and the Friedman lab, looking at actual clinical serum, screening those as well. So more with a tilt towards maybe developing effective immunogens or vaccines. Those experiments were developing a set of reference antibodies that we have really detailed binning and mapping data for, and then using those to probe preclinical CRO that might be infected or vaccinated and really understand what are the immune responses in those samples. So that's probably one of the other facets that we've seen the LSA being used for, in these sort of COVID 19 or general vaccine type assays.


0:40:19.3 John: Another question. You showed categorical data being incorporated into the network pods. Can numerical or sequence data be added?


0:40:29.3 Noah: Yes, yeah, good question. So in the... Excuse me, orthogonal data option where we can add it, we did have, in the example I had showed, it was categorical or kind of a certain light chain or you're mapping to a certain domain. But we could include numerical data. One of the best examples of numerical data would be affinities. So we can add affinity values in there, and then represent that as a heatmap scale on the nodes within the network plots. So that's great, because I kind of alluded to this in the top that, if you do amend on a certain community or bin, you have a bunch of constituents that are hagen epitope you want, but which one do you choose? Within that, you can see the range of affinities and sort of easily visually pull out the ones that are most interesting to you, for example.


0:41:15.8 Noah: Sequence-wise, there's definitely the ability to add sequence in, as orthogonal data, and you might have seen it in the gear menu, where I was kinda zooming in there and making changes to the network plot colors, there is an option to label the data as well. So the nodes, instead of coloring them based on a different property, we can actually put a little call out to a label next to them. So we do have some customers that use sequence data and represent that around the different clones in the binning network plot, to show the relationship between a given sequence and where they fall out, in terms of competitive binning.


Can I combine and analyze several smaller binning data sets?


0:41:55.5 John: Another question. Can I combine and analyze several smaller binning data sets?


0:42:01.9 Noah: Yes. If you did have a scenario... If you had a scenario, where you were maybe running acute smaller experiments or you had a delay, maybe you wanted go back and bin more and combine the data sets to analyze all together, there is a merge function in the software. So you can go ahead, easily take two independent binning experiments and merge them together to one master file that you can analyze in unison, to really simplify kind of those comparisons with the two separate data sets.


0:42:37.6 John: Okay. We have time for one last question. We'll make sure to address all of the other questions that were asked. But our last question is: How do you competitively bin with multi-binning antigens?


0:42:52.9 Noah: Okay, yeah. So there's sort of two ways that binning is most effective on the LSA. People that are familiar with competitive binning, there's the classical sandwich style format, which is primarily what I showed in this data set and what was used for the evaluation of the OmniChicken platform. The alternative format is the pre-mix format, where instead of running a sequential injection of antigen followed by a clone and solution, to see if a tri-molecular complex forms sequentially, we actually pre-mix the clone with the antigen and run that mixture across the surface simultaneously. So, that tells us the same answer in terms of whether a complex forms, it allows the binding sites for multi-down antigens to be saturated. So in other words, if you have a multi-down antigen, you have maybe a homodimer, you have two epitopes on there, that can be recognized. We can't flow that in a classical assay because there is potential for one of the sides of the dimer to be exposed, to have reactivity to the analyte clone. So we basically saturate both sides by incubating a molar excess of the antibody with the antigen, and then flowing that mixture across, in order to really determine if there truly is unique epitope recognition or not, between the surface bound antibody and the antibody solution. Good question.


0:44:16.8 John: Well, great. Thank you so much, Noah. We'll make sure to answer everybody else's questions that we didn't get time to answer to today. And also, we'll be hosting a on-demand version of this webinar on our website, for you to come back to and share with your colleagues as well. I really wanna thank you, Noah, for a really great presentation, and thank all of you for attending. Please make sure to check back in with us next Thursday, at the same time, for our Kinetics webinar with Dan Bedinger, and you'll be receiving an invite for that shortly. Thank you all so much for attending. Thanks, Noah.


0:45:00.0 Noah: Thank you, everyone. Bye-Bye.