De-Risking The Pipeline With New Live-Cell Insights

[0:00] All right, so by show of hands, just to start this off, how many people here are familiar with image-based profiling? 

Cool, so we have maybe two, three people. And who is familiar has used molecular profiling methods before? RNA, sequencing, proteomics, metabolomics, all right, so we're familiar with that. 

So, for those who are not familiar with image based profiling, in the next 10 minutes or so, you'll learn why so many drug discovery groups have incorporated image based profiling into their high throughput screening workflows, and why the live-cell version has become the method of choice for them. 

I'm Felix. I'm co founder at Saguaro, as I've been introduced. We’ve helped, at this point about 200 drug discovery groups get better drug insights with our non toxic dyes, but also with our live cell screening and drug profiling service. 

All right, so let's start this off with the Drug Discovery Data Challenge. What's at the heart of the challenge we have with drug discovery data? And I don't think it's a secret to anyone here at this point, but the data that we use to make the critical drug discovery decisions that we do in our work often relies on data that is of suboptimal quality, and the first part of this problem is that the assays that we rely on are simply oversimplified.  

Whether we use a target-based approach or a phenotypic based approach, we often rely on a uni dimensional view of biology. We simplify the models where we look at one or two specific biomarkers that represent the disease of interest. So both methods have their advantages, but in both cases, considering that we're looking at simple, unidimensional view, whether it's molecular target, like a protein or an enzyme, or specific phenotype that represents a predefined outcome that we measure, like cell viability, could be other things, like cell number, even like lipid droplet count, or micronuclei count. These things are still predefined and fail to capture the complexity of biology. 

Second part of the problem is, well, we do have assays that are data rich, that provide a high density of biological information and can help us better understand or have a fuller picture of drug response. And these are the omics, the molecular profiling methods that I was asking about when starting this talk. 

And they're pretty good actually. They give a more comprehensive view of cell response, but they lack scalability, and even taken on their own, they look at a specific layer of the functional genomics machinery, and we could say that taken alone, they could fail to capture a more holistic view. It doesn't mean that they don't have very good advantages. They each have their own advantages. 

But that's funny, because I was going to say that to summarise these advantages, I wanted to cite an author that is actually in the room today and his fellow authors in a 2022, paper on phenotypic drug discovery. 

 Mr. Mehsan and his fellow co authors who explain how scientists are able to quote, unquote, with these types of assays evolve from serendipity to a structure activity relationship based approach that can minimize safety risks while optimising phenotypic activity to increase the chances of clinical success.  

So in other words, again, these technologies are pretty powerful. They can provide clues to a drug's mechanism of action, but although they're they provide a high density of biological information, like I said, they still lack scalability, mainly because they're costly to run in high throughput screens and taken on their own. As I mentioned, they'll they fail to capture what we could define as an ultimate manifestation of disease or drug effect.  

They will fail to capture cells in context because they're destructive or could require fixation. So we cannot probe cells in their spatial and temporal context, basically. 

So this is just to summarise the last two slides, the two part of the problem, which can be summarised into a simple trade off between quality and convenience, where we can decide to have data rich complex assays that can provide data that is of quality, but that's at the expense of achieving the higher scales that are often more compatible with higher throughput screening assays.  

And in that context of having scalable assays that are more convenient, they tend to be oversimplified. It doesn't mean it's bad, right? Because it's all about trade offs. It allows us, again, to use in higher throughput. But the question today is more like, what if? What if we could use a method that captures the complexity and the richness of biological response, all the while achieving the convenience and scalability of other methods. 

So, I'm looking at the time I need to accelerate quite a bit. But bottom line, I mean, you might have guessed at this point, image-based profiling is a pretty powerful method. It transforms images into well, rich microscopy images, into meaningful drug insights. So we build image profiles. Instead of building profiles or lists of molecules where they're enriched with their enrichment levels, like proteins or RNA, we use image features to build these rich profiles, and it's surprisingly powerful.  

I'll go pretty quick through these, but basically it just shows that with image based profiling, we can provide not only overlapping information about biological response, but actually complementary information that's not captured with alternatives. And it's pretty scalable, much cheaper. It opens the door to discovering new mechanism of action, as been shown in the literature. I have some examples I would have wanted to tell you a little bit more about. 

But there are still limitations with these around fixing cells, not looking at cells in their native context. So although it's great, there's still improvements that are possible to be made, and in live cells, we can theoretically probe cells in their more native environment, have a better representation of cells in context. So sorry, quick parenthesis. Like the talk, like started a few minutes after my time, right? Just to be sure. 

Okay, all right, okay, thanks. 

So, yeah, coming back to the talk, basically with live cell image based profiling, we're seeing that with the right assay technology on the left here, there's a potential to even potentially outperform cell painting in terms of the phenotypic profiling performance, and I'll come back to this in a moment.  

But also with a technology that's completely nontoxic, and even better, like doesn't affect gene expression patterns, we could be looking at something that probes cells in their true native environment, which ultimately is what we want to do, right?  

We want to capture cells and get the most accurate biological data as possible, which combined together those first two aspects would lead us to potentially reach new kinetic insights that we wouldn't be able to capture otherwise. 

But again, like there are limitations, we don't necessarily have assays, sorry, we don't necessarily have assays that are data rich for live cells, meaning data rich but that are non toxic, and there's this new set of logistics required in high throughput screening for working with live cells, and that's what we built with ChromaLIVE. 

We realized that there were these challenges, and there was this potential for basically making image based profiling, which is already very powerful method, even better. And basically we developed this non toxic data rich dye that allows us to, today, we can say, provide the richest drug annotations in image based profiling.  

The data I was showing earlier was actually a direct comparison with cell painting, showing that ChromaLIVE basically provides a better performance based on mean average precision scores in reference libraries to basically profile cell response, and we've recently initiated work where we screen thousands of compounds in live cell painting in collaboration with Merck in Boston, where we develop new workflows.  

And by the way, they're not that complicated, it's just that we put them into a how to guide so that people can understand how to perform these types of assays. It's pretty straightforward, but there are still specific considerations to take into account, and we've made an effort into that. 

So basically, these are other examples where ChromaLIVE has been used. It's pretty good in 3D cultures, but yeah, if you want to learn more we have a booth in the startup zone. Come see us, or even if it's not your interest specifically dyes, we're always interested to talk about how to improve the quality of our drug discovery data. Thank you for listening.