At the 2024 BioFuture conference held in October in New York City, GEN sat down with Xiaorong He, PhD, Senior Vice President, Site Head Development U.S. and Head of Global Development Sciences, Boehringer Ingelheim, to discuss her perspective on applications of artificial intelligence in pharma. During the meeting, He participated in a panel that explored some of the evolving uses of AI within the pharmaceutical and biotechnology industry, and the potential impact of the technology on various applications in the space from drug discovery to clinical trials.
During the conversation with GEN, He highlighted several investments that Boehringer Ingelheim has made in AI-based tech. About a year ago, the company announced a multi-target drug discovery collaboration with Ten63 Therapeutics. Ten63’s computational drug discovery engine uses generative AI and physics-based models to find drugs for difficult targets. Also last year, Boehringer announced an agreement with Phenomic AI to find important targets in stroma-rich cancers. The company’s scTx platform uses advanced AI and machine learning algorithms to integrate and analyze single-cell RNA datasets. But Boehringer’s investment in AI extends beyond drug discovery to include tools that support various activities across its value chain. This includes partnerships with Google Quantum AI and IBM.
Senior Vice President, Site Head Development U.S. and Head of Global Development Sciences, Boehringer Ingelheim
He discussed these partnerships and more during her conversation with GEN. Prior to joining Boehringer Ingelheim in 2010, He had stints at other large pharmaceutical companies including Pfizer and GlaxoSmithKline, and worked briefly at a startup in China.
What follows is a version of the interview that has been edited for length and clarity.
GEN: Tell us a bit about your role at Boehringer Ingelheim and what it involves.
I have two hats. One is as a site head for development in the United States, which is a 450-person organization [focused on] non-clinical development. [Getting] a molecule from research to launch is about a 10-year period and so we take care of anything non-clinical. We make sure that there is a drug product and [work on things like] safety.
The second hat is in development science, which is more of the innovation arm. [We focus on] the overarching topics that are really important for Boehringer such as sustainability, patient centricity, and of course technology innovation, and also scientific excellence engagement. We have a global team in Japan, China, Germany, and the United States.
GEN: At BioFuture, you were part of a panel discussing AI in the pharmaceutical industry. Let’s talk about where AI is currently being used more broadly.
AI has been around for more than 10 years. Machine learning and predictive modeling started 20–30 years ago but now it’s everywhere. You can find use cases in the entire value chain starting with discovery all the way through to the commercial space. At Boehringer, we are truly embracing this. In the innovation unit, which has research, non-clinical development, and early clinical trial design, we’re looking at how you use AI to identify drug targets and molecule design, both for small molecules and antibodies.
And then in drug development, safety assessment is one of the main things because toxicity is one of the two leading causes of drug attrition. We are looking at how we can use AI to help streamline safety assessment and make it more accurate and faster. It can really help if we are able to get this right, and we are hoping that we can enhance our success rate. Our ambition is by at least 50% through digital innovation by 2035.
GEN: How about at Boehringer Ingelheim specifically?
I’ll give you an example which is already implemented. ADAM [Advanced Design Assistant for Molecules] is a tool we use for molecule design in the small molecule space to generate new structures to guide medicinal chemistry. This has been implemented for seven years and people are using it to identify better structures with better properties, developability, and so on.
In development, we are talking about digital twins. In drug development, you have two big chunks: the drug substance/drug product manufacturing and the pharm tox. Traditionally, when you are talking about formulation development or process development, you use a lot of resources and make batches from grams to kilograms. Having the digital twin predictive modeling can really help to assess if this is the right formulation. And this saves a lot of time. So those are two examples.
The other area for AI, which could be seen as low-hanging fruit but is sometimes very difficult to realize at scale, is productivity gain. Tools like Microsoft Copilot help with document generation. Now we have a platform called IQNow which is an AI-powered knowledge management platform that is rolled out company-wide. We have maybe 30,000 users. Since the implementation about three or four years ago, the time savings, as of a couple of weeks ago, is 116 years. It’s pretty amazing.
GEN: You mentioned earlier that there are a lot of areas in the drug discovery and development space where AI could be used and is being used. Is the data available to train the models to really help us in that regard?
That is the billion-dollar question. We have the data, but the caveat is the data is everywhere. And that’s a general problem, not just a Boehringer Ingelheim problem. We have so many legacy systems and the data is trapped in the different systems which don’t talk to each other. Now the challenge or the opportunity is how do you link those systems to free up the data. [At Boehringer Ingelheim, we] put them in a data lake, what we call BI Data Land. So, we are slowly putting data into the BI Data Land. Everybody’s supposed to do that, not just R&D, so that eventually everyone has it to do the data mining, analytics, [and] to train models.
Even though we’re not quite there yet, with the dataset we have right now, we have some specific use cases where we have access to data. And you’d be amazed actually how much you can do even when you don’t have everything sorted out. To get highly structured curative data is going to take years and millions of dollars. But actually we don’t have to do that. Because it’s about how we are using AI to answer specific questions. It’s the context of use. For example, for [drug] safety, there are so many aspects we can tackle. Sometimes, we can use unstructured data, for example, publicly available FDA data [on] drug labeling, approved drugs, doses, and so on. You can tap into a lot of information which can help with the training.
GEN: And I would imagine that as you get insights from that data, you can also feed those back into your models to further improve them.
Yes. And the important thing is leveraging the power of lab partnerships. For example with the FDA, we had a collaborative agreement [that used] a federated learning approach. We don’t share our confidential information, they don’t share theirs, but we’re sharing experience about models. We can tap into each other’s datasets and see how we can advance the safety assessment using AI-powered tools.
GEN: Speaking of partnerships, there were announcements in previous years about partnerships with IBM and Google QuantumAI. Are those still in play?
The IBM partnership is fairly recent. We are using the IBM partnership to look at how we use AI to do antibody drug discovery and for predicting probability properties. And then with Google, everybody loves AlphaFold. But there are other partnerships in the drug discovery space—[for example] we have Ten63 and Phenomic AI. There are quite a few.
Another thing is that we are also leveraging the power of a consortium in the pre-competitive space. These are companies that join together to say what are the things that we can share with each other, and then use AI to advance the cause of patients. An example is virtual controls used in preclinical studies. If you share the control study data, this is not sharing any private company information. Imagine having control study results from 20 companies. Maybe in the future, we can do virtual control using [an] AI-powered model. You’d save a lot of resources.
GEN: It sounds like time savings might be the major impact of applying AI?
We’ve seen isolated cases like when we talk about digital twins [for example]. Traditionally, it’s going to take kilos of APIs [active pharmaceutical ingredients] and six months of time to develop a process. If I can use predictive tools, instead of doing 20 experiments at different scales, I do two and then predict and confirm. You can quantify how much APIs we’re saving and that relates also to the time saving because two experiments are a whole lot faster than twenty.
Really the holy grail for tremendous savings potential is the probability of success. And this is something that we’re embarking on right now. So we have a strategic initiative called Computational Innovation Alliance. Basically, we’re saying how we can use AI to enhance our probability of success. But there are already examples that generate savings like IQNow and ADAM, making us confident that this is the right area to invest.
GEN: What do you think the future of AI in pharma will be in the next 5–10 years?
I think a lot is going to happen. Just in the past year and a half, many things have already happened. AI is one of the very important tools to help us enhance productivity in pharma in general in terms of probability, speed, and resource utilization. We know that the success rate in the pharma industry is single digits which is really not that great. It’s very difficult because diseases are complex. It’s not easy to find something that [works] and it also has to beat the standard of care. The bar is getting higher and higher and it’s going to be more and more difficult to get new drugs on that market that beat that higher standard of care. I think that AI can really help us in that space. I see people talking about virtual clinical trials, virtual manufacturing, automation, and drug design powered by AI. There are very successful examples out there and there’s probably going to be more of that type of thing happening in 10 years.
GEN: Last question. What are some initiatives at Boehringer Ingelheim that you are excited about that you can discuss publicly?
We’ve talked about AI and how we can use it to enhance our success rate. There’s a big strategic investment there. The second is about sustainability. We are very committed to sustainable development for generations. We have a very strong focus on health equity and access for patients, and in development, [we are]looking at the whole life cycle from starting materials all the way to packaging. Our ambition is by 2030, 100% of our pipelines are going to use the equal design principle.
Another area for us is patient-centricity. We are actively engaging patients to get their insights early to guide our development effort. We are developing an AI-powered digital app to get their input earlier and then we can effectively incorporate it in our design. We are using the app not only in development but also in clinical trial design and with the global patient engagement network.
