Erik Ingelsson (Senior Vice President of Genomic Sciences at GlaxoSmithKline) joins us to discuss the importance of genetics in drug discovery and the importance of diversity and inclusion in the drug discovery process.
Please note the transcript has been edited for brevity and clarity.
FLG: Hello, and welcome to the latest “A spotlight on” interview. Today, we’re going to be talking about genetics and drug discovery. I’m joined by Erik Ingelsson. Erik, please could you introduce yourself and tell everyone a little about what you do?
Erik Ingelsson: Sure, and thanks for inviting me to this interview. I’m Erik Ingelsson, and I’m Senior Vice President of Genomic Sciences at GSK. Before joining GSK, which was about two and a half years ago, I spent around 20 years in academia – both in Europe and the US. My last position was as a Professor of Medicine at Stanford University.
At GSK, I lead a department of about 300 scientists. We are working on a range of different methods in genomics, covering human genetics, bioinformatics, computational biology, genome biology (including both CRISPR and other gene editing tools), omics (both single cell and bulk-type sequencing approaches, and mass spec approaches). Obviously, this has a really broad set of applications within genomics. We work on everything from early target identification and validation, all the way through the whole pipeline.
FLG: Thank you. To dive straight in, what are some of the current challenges in drug discovery and development? What are our success rates in bringing something to patients?
Erik Ingelsson: The main challenge in drug discovery and development is the low success rate. If you look at new molecules entering clinical trials, we’re at about a 10% success rate. This means that only 10% of medicines that start with a phase I trial reach the market, get approved, and get to patients. If you look even earlier in the pipeline, for example at new programmes where we have a target and try to find a corresponding medicine, it’s even lower – less than a 1% success rate.
I’d say this is the biggest challenge in drug discovery and development, though there are obviously additional challenges. One of the reasons I think this is an issue is that we’ve been relying too heavily on correlations. We’ve focused on things like biomarkers or gene expression levels in patients versus non-patients, which are correlations, but are not necessarily causation. We’ve also relied quite heavily on in vivo models, like mouse models, in drug discovery.
FLG: What are the current timelines and costs associated with bringing a drug to patients?
Erik Ingelsson: The timelines are typically around 15 years or more, though this depends a lot on the type of drug and the disease. For certain diseases, the randomised clinical trials will be very large and long. For others, like rare diseases (specifically orphan drugs) or in oncology, you will have a different path, so it varies a little bit. It also depends on where you start. If you have more knowledge about the biology, you might be lucky enough to have a medicine already developed for some other indication that can be used here – this would shorten your timeline. But typically, I would say at least 10 to 15 years – and sometimes longer.
In terms of costs, a longer timeline obviously also drives the costs up. The combination of the long process to bring a medicine to patients, plus the low efficacy and high failure rates – those factors are basically what’s driving these high costs.
There are various numbers out there, but generally the costs are high. Recent numbers I’ve seen from scientific publications are in the range of 1 billion to 3 billion US dollars per drug that reaches market. And that’s when you’re taking expenditures for all the failures into account as well. That’s clearly an unsustainable situation, and it’s driven by the factors I just mentioned – the long process and lots of failures. Especially since many of the failures happen when you get into phases II and III, and these are the most expensive parts of the drug development process.
FLG: You talked about mouse models and the challenges they’re presenting as well. What have our historical approaches been? And what are the growing and emerging options?
Erik Ingelsson: For target-based discovery – where we start with a specific target and try to find medicines for it – we mostly start with known biology. So, this could be known pathways or other biology that has already been disentangled in great detail and is quite anchored on animal studies. We would also focus on target classes that we know we can find molecules for, such as kinases, GPCRs, or certain drug classes. So that’s one approach. Another approach that has been used a lot is phenotypic screens. And that’s basically an approach where you do a molecule screen and try to find new molecular entities with the ability to alter a cell phenotype. You’re not really looking for a specific target, you’re looking more for a change in the phenotype. And you typically would use cell-based or animal models for that.
Those two approaches have been the traditional ways of working. Nowadays, a lot of pharmaceutical companies are pivoting towards having a stronger focus on human genetics, computational biology, functional omics, and artificial intelligence (machine learning), as the starting point. The idea behind this is based on a couple of things. One thing is that we’re trying to expand into biology that hasn’t been looked at before. Without the constraints of what’s already known, scientists can generate more hypotheses or look at hypothesis-free approaches – this was the big change in genetics 15 years ago when genome-wide association studies came into practice. Also, we’ve been able to focus on humans more strongly as the model system for target identification and validation. Rather than doubling down on mouse studies, GSK and a lot of other companies are focusing much more strongly on humans, meaning a lot of human genetics is now the starting point.
FLG: When it comes to identifying a drug candidate, there’s obviously a myriad of factors going on. What makes an ideal drug candidate?
Erik Ingelsson: I think the most important thing is that it addresses an unmet need. It should be a drug for a disease where we currently don’t have good medicines, where there are patients in need, and where these patients are not being treated optimally. That’s the main driving factor, I would say. But then there are obviously a lot of different ways to try to get to that.
I think in terms of other factors, you need to have a strong biological connection, primarily from human systems. So that’s where genetics is really helpful. You need a very strong reason to believe that this target is involved in either the development or the progression of the disease, pointing towards causality.
A third important factor is that you should have a feasible path to a medicine. For example, you should believe that this particular protein can be drugged – that it’s tractable and that there is some way of modulating it. There should also be a translational path – when we discover a molecule, we should have different types of biomarkers for target engagement and pharmacological effect that we can measure to ensure it is having an effect.
Other factors would be a lack of safety signals – we should have no indication that this drug would be dangerous in our preclinical systems. And then obviously, also there should be some novelty there. For a company to develop something, it should be a new type of medicine heading a new type of pathway – otherwise it wouldn’t really benefit patients.
FLG: What are some of the key benefits of a genetics-first approach to drug discovery?
Erik Ingelsson: By working with genetics, we can get indications of a drug’s causality in terms of disease development and/or progression. By using human systems, we also get around the problems of correlation in biomarker studies or mouse studies. If you think about this schematically, we can compare a randomised clinical trial (the ultimate trial we need to do before registering a drug) with, for example, a genetics study. They have a lot of similarities; you start with your eligible population, and in the randomised clinical trial, you randomise individuals into either getting a drug that (for example) lowers blood pressure, or a placebo. Preferably you would do this blindly. And then you follow, in this case, the blood pressure changes over time, and also the progression of cardiovascular disease.
Now, with a genetic study, you have a similar situation. You have the randomization that happens at the conception, and then you’re comparing the blood pressure lowering variant with a neutral variant (the placebo) and then you follow blood pressure changes and cardiovascular disease over time. It’s conceptually attractive, and that’s why some of these types of genetic studies are called Mendelian randomization – Mendelian meaning having happened at conception and then randomization of the alleles. So that’s kind of the basics of why human genetics can model and predict causality in a way that other methods really can’t.
FLG: Fantastic. What other genetic evidence do we need to consider in the context of drug discovery?
Erik Ingelsson: What we really mean by genetic evidence is that we have evidence of a causal gene. So, when we start with genetics, we typically have a statistical association between the region in the genome and a disease or disease-relevant phenotype, such as blood pressure, LDL, or some other type of biomarker that is relevant for the disease in question. We typically mean that it’s significant, so taking all the multiple tests into account, we can say it has genome-wide significance, that it is replicable, and it is robust. But that’s not enough; that only points us to that region. Something in that region is telling us that there is a causal association between the region and disease. Now, we need to connect the dots.
The first step is to connect the dots between genetic variation and a single gene. And that’s what we call variant to gene studies (V to G). When we talk about what genetic evidence is, we’re really talking about the strength of the evidence of a causal gene in the region and the knowledge about which gene that is. Because that’s where we need to start our drug discovery efforts – we need to understand a specific target. It’s typically not enough for us to know that there is something in the region, we need to know exactly what it is in the region. That’s the typical starting point. Additionally, it’s also beneficial for us if we know about the directionality. So, for this gene, is upregulation or downregulation going to be what we’re looking for? Are we looking for an agonist or an antagonist? That’s also something very important for us in the early steps.
So, I would say a lot of the main difficulties right now in translating human genetics into drugs are related to this variant to gene mapping. From the big number of associations that we’re aware of, I would say that maybe in up to 15-20% of cases we can be very confident about the causal gene – it’s either a coding variant, or has a very strong colocalisation, or there are other approaches that allow us to be very confident that it’s a causal gene. In the rest, so 75 to 80% of cases, there is either no evidence for a causal gene in the region, or it points to several and we’re not really sure.
This is a big challenge right now in the field, but sequencing can help. We’re working a lot with both exome and whole genome sequencing. But it’s not the sudden solution to every challenge, because a lot of the variation is actually due to the fact a lot of the signals are going to be related to regulatory elements in the genome. So, it will be necessary to work with other things as well.
The other big obstacle typically would be the directionality. So again, before we can start the drug development programme, we need to know if we’re looking for an antagonist or an agonist. And even with sequencing, quite often it won’t help, we can have some evidence or some strong beliefs that this is a loss of function variant, and that’s good, but quite often, it’s a missense variant, so we don’t really know the directionality. We would have to work with functional experiments to figure that out.
Those are probably the main things in terms of challenges related to genetics. There are also a lot of other challenges in the early stages that are more related to the first experimental approaches. For example: Do we have the right model systems? Do we have a good in vitro model that reflects the disease? A lot of these hits that we’re finding from genetics as well are very novel in terms of biology. They might have totally unknown pathways, and unprecedented variants. So that is also tricky. We are working through some of these challenges as well.
FLG: Another challenge is that historically, clinical trials have often lacked diversity, especially in regards to sex and ethnicity, which impacts patient outcomes. What do you think we need to do to ensure our future endeavours are representative and unbiased?
Erik Ingelsson: This is a very important issue for GSK generally, and also for my department. We have a very ambitious DEI (diversity, equity and inclusion) agenda. This centres around different aspects either related to people (things like recruitment, development of diverse talent, and good representation), or business (things like health equity, inclusive research, access to health care, and clinical trials that reflect different populations, minorities, and communities). This work is all about expanding scientific opportunities in underrepresented communities to level the playing field. So, we have a very broad agenda.
In genomics specifically, we’re working on a whole range of projects that hit all these different pillars. This work ranges from trying to get a more diverse workforce – we’re working very hard on that and it’s something that is very important to me – but we are also forming several collaborations with scientific organisations and educational organisations, such as historically black medical colleges. The aim of this work is to broaden genetics to ensure underrepresented minorities are included at every stage. And there are several big efforts ongoing in these areas that I hope we are able to launch within the next year.
FLG: If we look to the future, what are you excited for in the next five to 10 years?
Erik Ingelsson: I am excited about the integration of human genetics and other approaches in genomics – so, computational biology, functional omics approaches, gene editing and sequencing, and machine learning approaches will be a step-change for how we think about drug discovery. We have already changed how we do this at GSK over the last one to two years, and I think it will lead to pretty dramatic steps in terms of how we link the genetic associations to causal targets and an understanding of the biology.
I’m also excited about increasing access to new datasets, both more diverse datasets, as we talked about already, but also datasets with deeper phenotyping, meaning we can look at the progression of phenotypes and subsets of diseases. Also, for me personally, I’m excited about expanding what we do with genomics beyond just target identification and validation and moving more towards having an impact on the whole pipeline – this is an area that is expanding quite dramatically right now in our work.
FLG: Lots to look forward to! Thank you so much for sharing. I’ve learned so much today. I’m excited to see what the future of drug discovery looks like and the impact of genomics. So thank you for joining me.
Erik Ingelsson: Thanks so much for inviting me! It was a great discussion. Thank you.
Watch more interviews in our series here.