Accelerating the Development of Genetic Medicines with AI

The advent of genetic medicines is enabling the development of therapies that can repair or replace a faulty genetic sequence underlying a disease. WhiteLab Genomics has developed an AI-based platform to enable target discovery and design of DNA and RNA therapies in silico and shorten development times. We spoke to David Del Bourgo, CEO of WhiteLab Genomics, about its AI-platform technology, the data it uses, and its role in a consortium to develop highly specific vectors for genetic medicines.

Daniel Levine: David, thanks for joining us.

David Del Bourgo: Thank you for having me, Danny.

Daniel Levine: We’re going to talk about cell and gene therapies, White Lab Genomics, and how it’s using its AI platform to improve the development of these advanced therapies. Perhaps we can start with the challenges of developing cell and gene therapies today. What would you say those are?

David Del Bourgo: Yeah, so as we probably know, cell and gene therapies are extremely promising treatments and some of them actually are in the market already. There are about 10 gene therapies in the market. However, the challenge is to be able to get more of those therapies to patients in need. Those are lifesaving therapies. They can really make a difference—like, for example, the one in spinal muscular atrophy. We can really cure kids from a really lethal disease. So I would say the challenge is they’re difficult to bring to the market. There are a lot of challenges in terms of potential toxicity, high dose, and also cost because they’re also difficult to produce today. So that would be the key challenges of gene therapy today.

Daniel Levine: What can artificial intelligence do to address those challenges?

David Del Bourgo: Well, if you think about the amount of data that’s available today in terms of genomics, proteomics, all the multi-omics we call, this is all information from biological data that has been turning into sequencing information. And the amount of data today is extremely large. It’s huge and it’s not possible for the human brain, even the most brilliant ones, to go and be able to analyze and cross-analyze this data. And this is where AI comes into play by being able to have machine learning approaches or deep learning approaches in order to analyze those data sets and extract the right information that’s going to be necessary to build an effective and safe gene therapy.

Daniel Levine: As I think about DNA and RNA therapies, it seems to me the challenges have less to do with target discovery and the targets because the targets of these therapies are generally well known and the challenges have more to do with various constituencies of a therapy beyond the DNA or RNA themselves—things like vectors and promoters and chemical modifications and backbones. Target discovery is part of what your system does. Can you explain the need with regards to target discovery versus these other elements of creating a genetic medicine?

David Del Bourgo: Yeah, actually it’s a very good point, but you’re going to see that our approach links the two together. So, our platform enables identification of highly specific biomarkers for targets. In that particular case, our Atlas platform enables identification of receptors that are highly specific to a cell type. So that will be the receptor where the vector will be, in the end, binding. And so, if we have highly specific, highly qualified receptors, that’s our target, we are then able to design properly the vector in a system that will be like a lock/key system. So if we have the right lock, we design the right key, and then we have an effective gene therapy or an effective vector delivery including delivery of the payload you mentioned. The payload needs to also be designed in a way that the expression of the desired effect is consistent and efficient.

Daniel Levine: White Lab Genomics has developed an AI platform to accelerate the development of gene and cell therapies. Give me a sense of what the platform does and how it works.

David Del Bourgo: So, the first step of the platform is, as I mentioned, is it’s a very exhaustive biomarker cellular atlas that’s going to be able to identify targets that are highly specific to a cell type. So, let’s say we want to look at cells in the eyes, retina for example, or cones. We’re going to be able to identify receptors that are very specific to those ones. And then we have another scoring system that we call Surface Score that’s going to score those receptors and qualify them being at the surface of the membrane, making them then the right target. That’s the first step. And then on the second step we’re going to work on with deep learning approaches on structural biology in order to design, as I mentioned earlier, the right key, the right protein components that are going to be able to bind to those receptors properly.

Daniel Levine: What are the data inputs and what are the outputs from the system? What do you see, and is there a process of iteration?

David Del Bourgo: So, the input is going to be a lot of multi-omic data—so genomics data, proteomics data, transcriptomics data, epigenetics data. A lot of them come from very well vetted public sources that we’ve been doing for the past five years. Then we add our own data sets so we can generate a new dataset and a new hypothesis in order to reinforce the models. So yes, there are iterations because we need to build the training set for the machine learning. And so therefore there are iterations until we have algorithms that are satisfactory. From a statistical standpoint, the output is going to be, I would say, it’s going to be sequences. So if we have to design a new AAV vector, the system is going to provide the sequences of those AAV vectors.

Daniel Levine: And how much of this is done in silico and at what point do you move from the computer to a lab experiment?

David Del Bourgo: Yeah, so the upstream part is very silico and what I’ve described is from the target discovery part until the design of the new vectors, this is done in silico, then we move to the wet part where we validate the predictions in vivo in order to validate them from a biological standpoint. So, there are iterations between wet and in silico.

Daniel Levine: What’s known about the benefits you provide? Can anything yet be said about the time and cost savings to getting a candidate into the clinic?

David Del Bourgo: Oh, getting a candidate into the clinic, yeah. So, there is our desire to divide the time, the nonclinical time from target discovery to IND by three. The earlier projects start on the smaller scope, right, because it’s like a step-by-step process and then it’s still early in the development. So, I think we see more a factor of two right now, but our target is to divide it by three.

Daniel Levine: And what’s been done today to validate the system?

David Del Bourgo: We have done multiple validations on different levels. The ones that we have validated, for example, are already validated experimental results that were already generated. We make the predictions and we check that this corresponds to the validated experiments. So, we’ve done that on multiple assets and also as we speak, we’re generating new biological in vivo data to validate new hypotheses. So I think in the next couple of months we’re going to be showing in different conferences those results.

Daniel Levine: I’m not clear on the company’s business model, it appears that you operate as a service provider rather than developing your own pipeline. Who’s the customer and how are you compensated?

David Del Bourgo: Yeah, we like to see ourselves more as a partner. And so, the way we collaborate is to enter into a collaboration agreement, R&D collaboration agreement, with our customers. Customers are biotech companies, pharma companies, as you saw in the WIDGeT Consortium, Sanofi, and including also academic labs. The way we’re compensated is through an upstream milestone royalties model. So, pretty common in our field because we do have, in many contracts, co-IP of the assets that are generated, although eventually it’s going to be the biotech and the pharma, I’ll call the pharma company that’s going to exploit it and bring them to the clinic, as you mentioned, and bring them to the market. Regarding internal pipeline, those are like projects we’re thinking about. It’s always in the background of our mind as we acquire more data and more knowledge. So we keep that in mind.

Daniel Levine: Given the modalities, it’s not surprising that you’re working a lot on monogenic diseases and cancer, but are there some examples of projects you’re working on you can offer?

David Del Bourgo: Yeah, we have a project that we’re working on—the metabolic disease in the liver with the Genethon Group. You mentioned, as you know in the press release regarding the widget consortium, we work on the AMD and podocytopathies that are responsible for kidney disease. We also work on a project on glioblastoma. So we are pretty agnostic in terms of therapeutic area, although we tend to have lots of projects in neuro and also more and more in cardio. But you see because our atlas, the Atlas we developed, covers more than 700 cell types. So, we are able to work on muscle cells or neuro or heart, different type of cells. And on that point we’re pretty agnostic.

Daniel Levine: You mentioned the WIDGeT Consortium. White Lab Genomics recently entered into a collaboration with Sanofi, the TaRGeT Laboratory at Nantes Université, and the Institut Imagine to launched the WIDGeT Consortium. Can you expand on what the WIDGeT Consortium is and how it’s funded?

David Del Bourgo: So, the WIDGeT Consortium is a consortium composed, as you mentioned, of Sanofi, the TaRGeT Laboratory from Nantes Unive, the Imagine Institut in Paris, which specializes in rare diseases, and White Lab Genomics. So, you can look at it if I make it simple, White Lab is the AI in silico specialist in genomic medicine, you have the two experts in the other pathology respective like AMD and podocytopathies, and then you have the pharma company, Sanofi. We’re going to be able to bring this candidate to the market. We’re funded as part of French initiatives because France 2030 is supporting France to be a champion in biotherapies and bioproduction. And so, this is an $18 million project with the support of the French government.

Daniel Levine: And what’s the goal of the consortium?

David Del Bourgo: So, the goal is really to create, using AI, a platform that’s able to develop highly specific vectors. And in that case, that’s why we created this consortium for two different therapeutic targets. And so, it’s just to show that they’re very different, to show that the platform that’s going to be developed is going to enable eventually to develop more AAVs, which is a challenge in the gene therapy field that are highly specific, that are nontoxic, and then can carry the right cargo to their targets. And so, this is the purpose of the consortium—to build this platform and to bring those two candidates also to the clinical phase.

Daniel Levine: This is a group of impressive partners and it’s, I imagine, non-dilutive funding for you that helps advance the platform. How validating is this deal and how significant is it to the growth of the company?

David Del Bourgo: Well, Danny, it’s very validating for us. We were founded in 2019, almost five years ago. We’ve been bootstrapping for many years and then we were lucky to get into Y Combinator and also get investors, raising a 10 million round last year. Now it’s time for us to scale and show what we’ve developed is of interest to large players such as Sanofi and also we’re having active discussions with other pharma companies. I think the fact that we announced Sanofi is also showing the other pharma companies [that] we are a credible player in the field. AI and genomic medicine can be difficult sometimes to understand what AI can bring in general to drug discovery or drug development. I think today a lot of companies have proven the impact of AI in biology. And so yeah, it is very validating and we believe it’s going to support our growth for the next years.

Daniel Levine: And how has the company been funded to date and how far will existing funding take you?

David Del Bourgo: So last year we raised 10 million euros in September 2022. This gives us visibility for three years. So we’re not in a rush to raise, but also as we want to grow faster and accelerate, we might look at raising again end of next year.

Daniel Levine: David Del Bourgo, CEO of White Lab Genomics. David, thanks so much for your time today.

David Del Bourgo: Thank you very much Danny. Appreciate it.

This transcript has been edited for clarity and readability.