When quality data meets pioneering GenAI for drug discovery

There’s a definite click between Elsevier and Iktos — the pioneer in generative AI for drug discovery. And this connection goes beyond the tech. Both companies share a love for customers and collaboration — and a distaste for smoke and mirrors.

The best models for the best data

Industry insiders took notice of the recent announcement abre em uma nova guia/janela of Elsevier’s multiyear partnership with the innovative French AI company Iktos abre em uma nova guia/janela to deliver an AI-driven synthetic chemistry platform for small molecule development. In short, the world’s largest and highest-quality chemistry database, Elsevier’s Reaxys abre em uma nova guia/janela, was opening itself up to Iktos’s cutting-edge AI for drug discovery. It’s a combination that will undoubtedly accelerate pharmaceutical companies’ Design-Make-Test-Analyze chemistry research cycle.

Elsevier is bringing its data, broad market appeal, user-friendly interface and application programming interfaces (APIs). Meanwhile, Iktos offers its best-in-class retrosynthesis AI technology, which has already proven itself with the world’s most important pharmaceutical companies.

With the agreement signed and announced, it was time to chat with the two main people behind the partnership: Iktos CEO and co-founder Yann Gaston-Mathé abre em uma nova guia/janela and Reaxys Innovation Director Dr Abhinav Kumar abre em uma nova guia/janela.

Streamlining drug discovery

Let’s jump in. How do you quickly describe your job to those outside the industry?

That’s a great analogy. Because like Google Maps, you can also potentially take different routes to your destination: the fast one, the cheap one, the most sustainable one, etcetera.

Yann: I concur. It is strange how people are often suspicious and skeptical about AI while also having unrealistic expectations regarding AI. It would be great if more people knew that it’s more about helping make better decisions — and how this will lead to improvements over time. AI is rarely the magic bullet. But of course, we are making real progress step by step. We are getting closer to dramatically changing the world of drug discovery.

And this is another thing I wish more people knew: We are in a great era of drug discovery, especially since AI in drug discovery poses no ethical issue of any kind compared to other areas. We are fortunate here. We don’t even work with patient data, so we don’t have to worry about privacy and data misuse issues. We essentially only work with data from chemical testing. And our task is straightforward: to help you develop better drugs more quickly for the benefit of patients. We are improving a process known to be very difficult, expensive and time-consuming.

Not an AI revolution but a tech evolution

Yann, you’ve said, “AI is not a revolution.”

Yann: Certainly, it’s not in our field of drug discovery. It's more of an acceleration of a trend that started around the 1980s with what was called computer-assisted drug design. Now, we are witnessing an acceleration because AI is becoming more available. It’s also an attitude shift: you can no longer ignore the fact that technology can help you be more efficient, faster, and so forth.

Abhinav: I'm always suspicious of new people entering the field of AI and overselling what technology can deliver — that whole ‘revolution’ aspect. This leads to unreasonable expectations and disappointments — which is a shame since technology can now bring something to the table.

How to stay real

So in terms of Iktos, what makes the company unique?

Abhinav: Let me take a crack at this — Yann will be too modest [laughter]. They have a very robust and innovative technology developed over years of accumulated industry experience working with many pharmaceutical companies. They understand the market since they come from that background. Secondly, what enables fast-paced development is a great collaborative team. Working with them as an external partner is already very collaborative and an absolute pleasure. I can only imagine what it’s like on the inside.

Yann, you’re blushing. Do you have anything to add?

Yann: I’m happy to blush [laughter]. My perception of what makes us unique is that we were one of the first companies to start working in AI for chemistry. In contrast to many other AI for drug discovery companies, who often wanted to demonstrate quickly what AI can deliver, we’ve been more focused on relentlessly improving the actual technology. Thanks to our many collaborations with numerous pharma companies, we understood the limitations and the ongoing need to improve this technology.

Our passion for science is really at the heart of the company. Yes, our technology is effective in accelerating drug discovery; we’ve demonstrated it on numerous occasions, and that’s important. At the same time, we do not like smoke and mirrors. We want to deliver on our promises. And the people who know us and have worked with us in the industry know we are very serious. I believe this significant difference will ultimately lead to better outcomes.

With tech, there's a huge reliance on academia. And they come up with brilliant ideas — and bless them for it. At the same time, you must keep an eye on the real ball: the real-world market. And Iktos’s eye seems sharper than most. Is there a secret?

Yann: It's not easy. You would like to try them all because there is so much fascinating and good stuff going on in academia. But of course, you must remain very objective — especially now when there’s still so much pressure to publish, publish, publish. Indeed, even the most brilliant ideas from academia do not always work in real life. So, this is why we've been working from the beginning on developing solutions that help chemists in real life.

Abhinav: Absolutely. And from Elsevier’s perspective, as a relatively large organization, we luckily get to work with some of the most innovative companies, and we listen to the big pharmaceutical companies we work with. At the same time, we invest a lot in R&D by partnering and collaborating with some of the most brilliant minds out there, like Professor Philippe Schwaller abre em uma nova guia/janela at EPFL abre em uma nova guia/janela, who was one of the pioneers of using transformer models, or Alexei Lapkin abre em uma nova guia/janela, Professor of Sustainable Reaction Engineering at the University of Cambridge. In this way, we get to be sort of a bridge between industry and academia and, in a way, cherry-pick — based on knowing the existence of a problem over here in the real world and then making a link with a researcher or research that can perhaps help us alleviate that problem.

Loving the work

Both of you ended up with pretty specialized jobs. Was there an initial spark or influence when you were younger that set you on the path to where you are today?

Yann: Certainly, for me. My grandfather abre em uma nova guia/janela, who died around 15 years ago, was a pioneering medical oncologist and famous in his time. He performed the first successful bone marrow transplant to cure leukemia. As a very important doctor and researcher, he influenced my decision to work in healthcare and for the pharma industry. It’s not random that I work in this field, albeit not directly as a doctor or researcher. I was driven by the inspiration I got from my grandfather.

Abhinav: I have a similar story. I come from a family of medics. My parents are doctors, and so are my brother and sister-in-law. So you are influenced by the conversations you hear around the dinner table. Secondly, I come from one of the poorest states in India, so I understand why you need drugs that are affordable and good. I think the combination of these two factors guided me to do a PhD in pharmaceutical sciences. As with Yann, I may not now be directly working in the industry, but we are both working on products that enable those discoveries.

In regard to your companies, what’s the greatest source of pride?

Yann: Well, when we started Iktos, there was nothing. And now, when I look back at what has been accomplished in seven years, I’m immensely proud. Because when you start a new adventure like that, there are thousands of reasons why it could fail. We started with a good basis: My co-founders are brilliant, and the new technology they had imagined was disruptive.

But what has really been the key is that we have been able to build a fantastic team that creates great technology, science and products within a culture of trust and fun. At the same time, it takes a lot of work to be the CEO of a startup company. It’s usually stressful. There are many moments when you would want to retire, have a permanent vacation, and sit on a beach drinking cocktails. But there are always more reasons to continue the adventure.

Abhinav: Again, I feel the same as Yann. If you ask anyone who leads teams, most will say that their source of pride is the cross-functional and brilliant team that they work with. What makes our teams particularly strong is that most of us are scientists. We’ve all been in the lab at some point, allowing us to walk in our customers' shoes. And this helps when developing new things.

I also feel really proud and lucky that I can work on cutting-edge technology. To take a research idea of pure mathematics and then develop, test and build it into something concrete our hundreds of thousands of users can use. There’s a creativity many don’t recognize: to go from conception to fruition.

Eye on the future

As we speak, you are gearing up to launch your first collaborative product — a significant milestone. How do you see the partnership evolving in the coming years?

Abhinav: Well, first things first. The upcoming product is still our focus: delivering it and enhancing it. The Iktos team has already done some additional great science, and we will work to incorporate that while making the overall enhancements required by any 1.0 version of a product. As for the medium and long term, there are other predictive models to build to accelerate the drug discovery process. We will explore other products we can build together in this area of de novo design and broader predictive chemistry.

Yann: I see it similarly. I hope in one year we can celebrate some very successful commercial achievements with several major pharmaceutical accounts that have been convinced of the value of our product. Early success is key to long-term success. Longer term, there are many opportunities to develop new technology by leveraging the fantastic Elsevier Reaxys data.

Picking the right horse

Where do you see the biggest challenge in maximizing this partnership’s potential?

Abhinav: Today, there’s potentially AI for everything. The challenge will be to work together to identify which of these many opportunities we’re willing to pursue. And I'm sure we will make a very well-informed decision together given the customer obsession and scientific expertise on both sides. But we must ensure we don’t get married to an idea. It’s always a challenge to detach yourself from something that seems very exciting. Look at how everyone is thinking about having large language model-based chatbots.

I was wondering how I was going to bring up LLMs.

Abhinav: LLMs may be exciting and good for a nice soundbite. But as Yann said earlier about smoke and mirrors, we need to take a step back and think: ‘What problem for the end user does it solve?’ Hopefully, this attitude will allow us to navigate these waters.

With all these changes afoot, is it necessary to build more flexibility into your processes beyond following an Agile work approach to be ready for the next potential game-changer, whether LLMs or another emerging technology?

Yann: I think it’s more about monitoring what’s happening. For instance, many companies and initiatives are developing LLMs for biological data, medical literature, etc. So, it will be fascinating to see how this develops and what benefits it will bring those working in those fields. Right now, I think the translation to chemistry is not so obvious.

Abhinav: I agree. And meanwhile in the world of synthesis, there remain many unsolved challenges. Together, we’ve solved two of them: how to make a molecule and evaluating whether it is feasible. But then there’s the chemist’s next question: What conditions should I use to maximize my yield? And then there’s the matter of scaling from these small-scale discoveries — for instance, you might require a whole new synthesis strategy, and it would be great to predict other routes around that.

We’ve returned to your Google Maps analogy of finding new routes to a greener future. The circle is round. Do you have anything to add?

Yann: I'm certainly very passionate about what more we can do together.

Abhinav: [Nods and smiles] Long live our shared journey.