During this year’s Discovery Week: Online, one of our most popular panel discussions was centred around Novel Drug Targeting. The discussion was moderated by Steve Rees (Vice President discovery biology at AstraZeneca), who was joined by Pieter Fokko Van Loo (Senior Director, Oncology – Immunology at Merus) and Vivek Vishnudas (Senior Vice President of Drug Discovery and platform Sciences at Berg Health). Due to the panel’s popularity and positive feedback from our delegates we have decided to provide the audio from the discussion as well as an edited, condensed transcript, which is available below.
Steve: Why is it important that we develop or identify new targets for the discovery of antibodies and antibody drug conjugates?
Pieter: There are a couple of molecule monoclonal antibodies in the clinic that have been great success, but there have also been many failures. We have some monoclonal antibodies that are potentially effective, but also toxic. We aim to find new targets or new target combinations that can overcome the toxicity, thereby creating a new therapeutic window. We typically try to use Microbial Limit Test platforms to find a combination of new targets that were created specifically. So, solving specificity is key.
Steve: So, this is all about identifying targets, which allow you to selectively target an antibody to a specific cell type to kill that cell without toxic effects elsewhere. Vivek, maybe you can tell us a little bit about how using artificial intelligence to find new targets?
Vivek: Yeah, thank you. So, our goal is to identify targets based on causal connections or causality. And what that does is allows us to identify a whole host of new targets, especially in an oncology setting, where resistance refractory becomes a major concern for patients who are receiving conventional treatment. Hitting that orthogonal mechanism on the orthogonal pathway is essential for us and identifying targets through the causal mechanism using Bayesian mathematics is the key.
Steve: And what sources do you use to identify new targets?
Vivek: We don’t many literature sources, we build our data internally. We build our datasets internally using omics datasets from human samples. But we do sometimes use the published data sources to overlay and benchmark our findings using AI.
Steve: Pieter Fokker, what methods do you use to identify new antibody targets?
Pieter: I think it’s a combination of both literature, databases, and database mining. The literature can tell you the type of mode of actions you’re after. But of course, knowing that our antibodies are expressed broadly in in lots of tissues, using databases and using an IAS approach is also helping us to identify targets combination. These are for a specific clinical indication, in our case, the cancer indication, whereas such combinations are not found in normal tissue.
Steve: Thank you, maybe I can touch a little on some of our work at AstraZeneca, as well. So, target discovery has become a major effort for us over the last three or four years. Across the pharmaceutical industry, we’re all trying to identify new drug targets, through which we may be able to intervene and better treat disease. We’re very heavy users of genomic data. The power of sequencing technologies, which now only costs a few $100 each per patient, allow us to perform analyses to identify new genes and disease. This is a very powerful route towards identifying new genetic associations, which in turn, bring new targets forward. Are you finding the same? And maybe do you have any examples where you’ve done this?
Vivek: Most recently, we have started using some synthetic lethality screening. using CRISPR, we’ve been able to identify some of these synthetic lethality partners for the new molecule that we have developed, that we take into IND. Functional genomics is a critical part of what we do. Of course, we also rely on proteomics analysis.
Pieter: Once you have identified the target of interest, you typically look into the expression profile, but also the expression levels of protein. If you want to achieve specificity, viability, you need to have the affinity of your fabs balanced to the expression levels. You also need to know what the expression levels are in normal tissue compared to the tumour tissue. Because of how we use patient samples, our expression levels are helping us also form a longer list of potential targets and filter down to a target that is going to be worked on.
Vivek: That’s an excellent point, we use quite a bit of prioritisation of targets internally as well, as Pieter was pointing out. A lot of it has to do with our AI model. In terms of picking the right target, certain connections that we have on the network are much stronger than others. So, we obviously tend to, drift towards those kinds of targets. But obviously, we look at expression levels between tissues, as well as the novelty around the targets. And what possible ways there are to drug that particular target; the drugging becomes a real challenge with some of these novel targets.
Steve: That’s a point I’d like to come back to in a second, but first I’d like to touch on a point that you both raised. This is the ability to access human tissue and characterise target expression levels in human tissue. Do you find it a challenge to access high quality human tissue to do these studies?
Pieter: It is doable, but you have to approach a couple of vendors and they have to look into the inventory to find a couple of samples which are specific for your question. So, the more specific it becomes, the more difficult it is. For example, if you require lung cancer cells from named resistant patients, for example, that is really difficult.
Vivek: Berg approach it a little differently; we took the human sample factor into account very early on in our discovery programmes. So, we were able to get into these collaborative partnerships with hospital centres and semi-academic type of research institutes that have access to a lot of these patient samples. One such example is our Department of Defence collaboration on breast cancer and prostate cancer, where we’ve been able to access the patient samples and well annotated patient samples very early on, for the same exact purpose that you just mentioned here.
Steve: Now and again, we find the same in our efforts, across therapy areas, whether it’s oncology, cardiovascular disease, or respiratory disease. Accessing patient samples is key to understanding what’s going on in a disease. Using techniques like genomics, transcriptomics, and proteomics is a huge effort for us. And I think that it will be for years to come.
But I would like also to go back to a point that made earlier. When we identify a novel target the clues in the name, it’s novel, quite often, there’s very little literature. So, would you like to share your experience of how your progress a novel target into a discovery project?
Vivek: One example is an E2 enzyme ubiquitin conjugating enzyme in the ubiquitin proteasome pathway. A lot of work has been done E1s and E3s, E2s are novel in many ways. They’re considered a challenging to drug because of the conserved active site. We had to look for allosteric pockets and allosteric designs. The real challenge was, which form of the protein do you actually drug and what is the pharmacologically most relevant form of the protein? We realise it was not the Apo protein, it was the ubiquitinated, the modified form of the E2 itself that was serving as a better target than the Apo protein. We had to use some human intelligence alongside AI as well, to find out which form of protein to go after.
Steve: Just to share a little of our experience, within AstraZeneca, we are working very hard to identify and validate our own targets. Novel is an interesting word; novel can be a new indication in an old target, all the way through to a target which may have less than 10 papers in the published literature. One consequence that we found, is as we’re moving increasingly into so called market target novelty, we’re moving away from what has traditionally been small molecule or antibody therapeutic space. And that has meant that our portfolio now consists of all sorts of different therapeutic modalities. Now, the hope, of course, is that with a range of different types of medicine, any target will theoretically be druggable. But that initial decision of trying to choose which therapeutic modality you take forward for a particular target is a major challenge.
Steve: What one or two technologies or approaches do you think will make the biggest difference in the future?
Vivek: I’ll take it first. So, I think for small molecule that the biggest challenge is going to be how to drug a particular target. The way we went about drugging a very novel target is using a fragment-based approach. We took a very systematic approach towards slowly growing the fragment. Just understanding what the pharmacological relevance of each of those steps were. And we were able to finally get to lead and lead op getting into candidate selection. So, I think, drug discovery, using a fragment-based approach really helped us with a novel target. That’s an approach I just wanted to highlight.
Steve: Maybe I’ll touch on that in a moment. But Pieter, would you like to comment first?
Pieter: Of course, one of challenges for novel targeting is definitely to having the right reagents available and making sure you understand the conformation of the of your targets. That’s where it starts, of course, in terms of regeneration of the antibody panel. Later on, we need to try to understand the biology because you also want to make sure that your drug works in the patient. So, investing into understanding the biology, and making sure you have the reagents to perform your discovery.
Steve: I think understanding the conformation of the protein that you wish to target, and availability of reagents is critical. And certainly, touching on something that you said around fragment-based discovery, I think we’re learning at AstraZeneca, that as we’re bringing forward these new targets into small molecule discovery, it’s affinity assays rather than some of the more classical biochemical assays that are crucial to run first. We need to try and identify molecules that bind, whether that’s fragment discovery or affinity, flexion, mass spectrometry, or DNA encoded library technology. These allow us to identify novel binding sites on proteins which we can then take forward. And at the very least, we’ve got a tool that we can begin to probe the biology with. So, we’re nearing the end of our time, any final comments?
Pieter: One final comment. I would always balance your portfolio between with more general targets. In this way, you leverage the risk of more challenging discovery, but high reward versus maybe easier targets. But of course, more people can do the same thing.
Steve: I think we would all agree with that comment. And so maybe just a couple of closing comments for me to wrap up. As we said, the identification and validation of novel targets, is critical to enabling us to deliver more efficacious medicines to patients and treat diseases in ways that we couldn’t do before. As you’ve heard from the panellists, key issues are the availability of human tissue to better understand disease and using data analysis methods and artificial intelligence to generate hypotheses. Moving on, the availability of reagents to then begin to progress projects are absolutely critical, something we have we haven’t really had time to discuss. Additionally, cell-based and animal model assays that are required to validate targets and build confidence.
So exciting times, exciting times as we bring novel targets forward. And with that, we will close this session. I would just like to offer my huge thanks to Pieter and to Vivek for joining this afternoon, and thanks to the organisers for arranging the session. Thank you very much.